diff --git a/make.conf b/make.conf deleted file mode 100644 index 223affd..0000000 --- a/make.conf +++ /dev/null @@ -1,47 +0,0 @@ -# These warnings indicate likely runtime problems with LTO, so promote them -# to errors. If a package fails to build with these, LTO should not be used there. -WARNING_FLAGS="-Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing" - -COMMON_FLAGS="-march=znver4 -O2 -pipe -fomit-frame-pointer -flto=7 ${WARNING_FLAGS}" -CFLAGS="${COMMON_FLAGS}" -CXXFLAGS="${COMMON_FLAGS}" -FCFLAGS="${COMMON_FLAGS}" -FFLAGS="${COMMON_FLAGS}" - -ACCEPT_LICENSE="@BINARY-REDISTRIBUTABLE" - -FEATURES="userfetch parallel-fetch parallel-install -ebuild-locks" -MAKEOPTS="-j12 --load-average 11.95" -EMERGE_DEFAULT_OPTS="--jobs 1024 --load-average 11.90" - -VIDEO_CARDS="amdgpu radeonsi" -INPUT_DEVICES="libinput wacom" - -AUD="pipewire pulseaudio screencast -alsa -gstreamer" -BLD="lto" -GFX="vaapi vulkan wayland -branding" -KIT="-gnome -gtk -gtk2 -gtk3 -gtk-doc -qt4 -qt5" -MSC="lm-sensors minizip sqlite -handbook -man -mysql" -NET="bluetooth -cups -ldap -wifi" -VID="v4l -vlc" - -USE="${AUD} ${BLD} ${GFX} ${KIT} ${MSC} ${NET} ${VID}" - -GRUB_PLATFORMS="efi-64" -GOPROXY="" -L10N="" - -DOTNET_TARGETS="net70" -PYTHON_TARGETS="python3_11" -PYTHON_SINGLE_TARGET=$PYTHON_TARGETS -LUA_TARGETS="lua5_4 luajit" -LUA_SINGLE_TARGET="luajit" - -# NOTE: This stage was built with the bindist Use flag enabled -PORTDIR="/var/db/repos/gentoo" -DISTDIR="/var/cache/distfiles" -PKGDIR="/var/cache/binpkgs" - -# This sets the language of build output to English. -# Please keep this setting intact when reporting bugs. -LC_MESSAGES=C diff --git a/make.conf/accept_license b/make.conf/accept_license new file mode 100644 index 0000000..bef28da --- /dev/null +++ b/make.conf/accept_license @@ -0,0 +1 @@ +ACCEPT_LICENSE="-* @BINARY-REDISTRIBUTABLE" diff --git a/make.conf/build b/make.conf/build new file mode 100644 index 0000000..6483cf3 --- /dev/null +++ b/make.conf/build @@ -0,0 +1,36 @@ +# These warnings indicate likely runtime problems with LTO, so promote them +# to errors. If a package fails to build with these, LTO should not be used there. +WARNING_FLAGS="-Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing" + +arch="znver4" + +procs="12" +make_load="11.95" +emerge_load="11.90" + +COMMON_FLAGS="-march=${arch} -O2 -pipe -fomit-frame-pointer -flto=7 ${WARNING_FLAGS}" +CFLAGS="${COMMON_FLAGS}" +CXXFLAGS="${COMMON_FLAGS}" +FCFLAGS="${COMMON_FLAGS}" +FFLAGS="${COMMON_FLAGS}" + +# Go shenanigans +CGO_FLAGS="${COMMON_FLAGS}" +CGO_CXXFLAGS="${COMMON_FLAGS}" +CGO_FFLAGS="${COMMON_FLAGS}" +GOPROXY="" +GOAMD64="v4" + +# Rust shenanigans +RUSTFLAGS="-C target-cpu=${arch}" + +FEATURES="userfetch parallel-fetch parallel-install -ebuild-locks" +MAKEOPTS="-j${procs} --load-average ${make_load}" +EMERGE_DEFAULT_OPTS="--jobs 1024 --load-average ${emerge_load}" +#EMERGE_DEFAULT_OPTS="--jobs 2 --load-average ${emerge_load}" + +PORTAGE_NICENESS="1" +PORTAGE_IONICE_COMMAND="ionice -c 1 -p \${PID}" + +VIDEO_CARDS="amdgpu radeonsi" +INPUT_DEVICES="libinput wacom" diff --git a/make.conf/misc b/make.conf/misc new file mode 100644 index 0000000..524f4d6 --- /dev/null +++ b/make.conf/misc @@ -0,0 +1,12 @@ +GRUB_PLATFORMS="efi-64" +L10N="" + +# NOTE: This stage was built with the bindist Use flag enabled +PORTDIR="/var/db/repos/gentoo" +DISTDIR="/var/cache/distfiles" +PKGDIR="/var/cache/binpkgs" +BINPKG_FORMAT="gpkg" + +# This sets the language of build output to English. +# Please keep this setting intact when reporting bugs. +LC_MESSAGES=C diff --git a/make.conf/targets b/make.conf/targets new file mode 100644 index 0000000..801b267 --- /dev/null +++ b/make.conf/targets @@ -0,0 +1,5 @@ +DOTNET_TARGETS="net70" +PYTHON_TARGETS="python3_12" +PYTHON_SINGLE_TARGET=$PYTHON_TARGETS +LUA_TARGETS="lua5_4 luajit" +LUA_SINGLE_TARGET="luajit" diff --git a/make.conf/use b/make.conf/use new file mode 100644 index 0000000..0d387c8 --- /dev/null +++ b/make.conf/use @@ -0,0 +1,9 @@ +AUD="pipewire pulseaudio screencast -alsa -gstreamer" +BLD="lto" +GFX="vaapi vulkan wayland -branding" +KIT="-gnome -gtk -gtk2 -gtk3 -gtk-doc -qt4 -qt5" +MSC="lm-sensors minizip sqlite -handbook -man -mysql" +NET="bluetooth -cups -ldap -wifi" +VID="v4l -vlc" + +USE="${AUD} ${BLD} ${GFX} ${KIT} ${MSC} ${NET} ${VID}" diff --git a/package.accept_keywords/itch-bin b/package.accept_keywords/itch-bin new file mode 100644 index 0000000..c10da74 --- /dev/null +++ b/package.accept_keywords/itch-bin @@ -0,0 +1 @@ +games-util/itch-bin::guru ~amd64 diff --git a/package.accept_keywords/libappindicator b/package.accept_keywords/libappindicator new file mode 100644 index 0000000..08c73e4 --- /dev/null +++ b/package.accept_keywords/libappindicator @@ -0,0 +1 @@ +dev-libs/libappindicator::steam-overlay ~amd64 diff --git a/package.accept_keywords/mangohud b/package.accept_keywords/mangohud new file mode 100644 index 0000000..4665ec6 --- /dev/null +++ b/package.accept_keywords/mangohud @@ -0,0 +1,2 @@ +games-util/mangohud::guru ~amd64 +media-libs/imgui::guru ~amd64 diff --git a/package.accept_keywords/obs-studio b/package.accept_keywords/obs-studio index 1d64898..0958626 100644 --- a/package.accept_keywords/obs-studio +++ b/package.accept_keywords/obs-studio @@ -1 +1,2 @@ +dev-libs/qr-code-generator ~amd64 media-video/obs-studio ~amd64 diff --git a/package.accept_keywords/pnpm-bin b/package.accept_keywords/pnpm-bin new file mode 100644 index 0000000..f477d9d --- /dev/null +++ b/package.accept_keywords/pnpm-bin @@ -0,0 +1 @@ +sys-apps/pnpm-bin::guru ~amd64 diff --git a/package.accept_keywords/protontricks b/package.accept_keywords/protontricks new file mode 100644 index 0000000..b248e20 --- /dev/null +++ b/package.accept_keywords/protontricks @@ -0,0 +1,2 @@ +app-emulation/protontricks ~amd64 +dev-python/vdf ~amd64 diff --git a/package.accept_keywords/vesktop-bin b/package.accept_keywords/vesktop-bin new file mode 100644 index 0000000..de52018 --- /dev/null +++ b/package.accept_keywords/vesktop-bin @@ -0,0 +1 @@ +net-im/vesktop-bin::guru ~amd64 diff --git a/package.accept_keywords/wine b/package.accept_keywords/wine new file mode 100644 index 0000000..d5194a1 --- /dev/null +++ b/package.accept_keywords/wine @@ -0,0 +1,3 @@ +app-emulation/wine-vanilla ~amd64 +app-emulation/wine ~amd64 +app-emulation/winetricks ~amd64 diff --git a/package.accept_keywords/xivlauncher b/package.accept_keywords/xivlauncher new file mode 100644 index 0000000..c7f00a7 --- /dev/null +++ b/package.accept_keywords/xivlauncher @@ -0,0 +1 @@ +games-util/xivlauncher::guru ~amd64 diff --git a/package.use/32-bit/plasma-meta b/package.use/32-bit/plasma-meta new file mode 100644 index 0000000..ed52f5d --- /dev/null +++ b/package.use/32-bit/plasma-meta @@ -0,0 +1,67 @@ +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=gui-libs/libdecor-0.2.2-r1 abi_x86_32 +# required by gui-libs/libdecor-0.2.2-r1::gentoo +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=x11-libs/pango-1.52.1 abi_x86_32 +# required by x11-libs/pango-1.52.1::gentoo +# required by gui-libs/libdecor-0.2.2-r1::gentoo +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=x11-libs/cairo-1.18.0 abi_x86_32 +# required by x11-libs/cairo-1.18.0::gentoo +# required by app-text/poppler-24.08.0::gentoo[cairo] +# required by kde-frameworks/kfilemetadata-6.6.0::gentoo[pdf] +# required by kde-apps/baloo-widgets-24.05.2::gentoo +# required by kde-apps/dolphin-24.05.2-r1::gentoo[semantic-desktop] +# required by kde-apps/dolphin (argument) +>=dev-libs/lzo-2.10 abi_x86_32 +# required by x11-libs/cairo-1.18.0::gentoo +# required by app-text/poppler-24.08.0::gentoo[cairo] +# required by kde-frameworks/kfilemetadata-6.6.0::gentoo[pdf] +# required by kde-apps/baloo-widgets-24.05.2::gentoo +# required by kde-apps/dolphin-24.05.2-r1::gentoo[semantic-desktop] +# required by kde-apps/dolphin (argument) +>=x11-libs/pixman-0.43.4 abi_x86_32 +# required by x11-libs/pango-1.52.1::gentoo +# required by gui-libs/libdecor-0.2.2-r1::gentoo +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=dev-libs/fribidi-1.0.13 abi_x86_32 +# required by x11-libs/pango-1.52.1::gentoo +# required by gui-libs/libdecor-0.2.2-r1::gentoo +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=media-libs/harfbuzz-9.0.0 abi_x86_32 +# required by x11-libs/pango-1.52.1::gentoo[X] +# required by gui-libs/libdecor-0.2.2-r1::gentoo +# required by media-libs/libsdl2-2.30.7-r1::gentoo[wayland] +# required by kde-plasma/plasma-desktop-6.1.5-r1::gentoo[sdl] +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=x11-libs/libXft-2.3.8 abi_x86_32 +# required by media-libs/harfbuzz-9.0.0::gentoo[graphite] +# required by dev-qt/qtbase-6.7.2-r5::gentoo[gui] +# required by kde-plasma/kinfocenter-6.1.5::gentoo +# required by kde-plasma/plasma-disks-6.1.5::gentoo +# required by kde-plasma/plasma-meta-6.1.5::gentoo[smart] +# required by plasma-meta (argument) +>=media-gfx/graphite2-1.3.14_p20210810-r3 abi_x86_32 +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=kde-plasma/breeze-6.1.5 qt5 +# required by kde-plasma/plasma-meta-6.1.5::gentoo +# required by plasma-meta (argument) +>=kde-plasma/plasma-integration-6.1.5 qt5 + diff --git a/package.use/ffmpeg b/package.use/ffmpeg index 3502f2d..1f291e2 100644 --- a/package.use/ffmpeg +++ b/package.use/ffmpeg @@ -1 +1 @@ -media-video/ffmpeg openh264 opus theora vaapi vpx webp x265 +media-video/ffmpeg libdrm openh264 opus theora vaapi vpx webp x265 diff --git a/package.use/gtk+ b/package.use/gtk+ new file mode 100644 index 0000000..ba3ae80 --- /dev/null +++ b/package.use/gtk+ @@ -0,0 +1,4 @@ +# required by games-util/itch-bin-26.1.9::guru +# required by itch-bin (argument) +>=x11-libs/gtk+-3.24.41-r1 cups + diff --git a/package.use/kimageframeworks b/package.use/kimageframeworks new file mode 100644 index 0000000..c9a7ebb --- /dev/null +++ b/package.use/kimageframeworks @@ -0,0 +1 @@ +kde-frameworks/kimageformats heif diff --git a/package.use/libdbusmenu b/package.use/libdbusmenu new file mode 100644 index 0000000..a160302 --- /dev/null +++ b/package.use/libdbusmenu @@ -0,0 +1,4 @@ +# required by dev-libs/libappindicator-12.10.1_p20200706-r1::steam-overlay +# required by @game-deps (argument) +>=dev-libs/libdbusmenu-16.04.0-r2 gtk + diff --git a/package.use/linux-firmware b/package.use/linux-firmware index de6347d..f68bc70 100644 --- a/package.use/linux-firmware +++ b/package.use/linux-firmware @@ -1 +1 @@ -sys-kernel/linux-firmware savedconfig +sys-kernel/linux-firmware savedconfig -initramfs diff --git a/package.use/obs-studio b/package.use/obs-studio index 50a30a8..c9d3609 100644 --- a/package.use/obs-studio +++ b/package.use/obs-studio @@ -1 +1 @@ -media-video/obs-studio browser +media-video/obs-studio browser websocket diff --git a/package.use/phonon b/package.use/phonon new file mode 100644 index 0000000..7a9b5fb --- /dev/null +++ b/package.use/phonon @@ -0,0 +1 @@ +media-libs/phonon minimal diff --git a/package.use/plasma-meta b/package.use/plasma-meta index 4dbc9ae..6110d57 100644 --- a/package.use/plasma-meta +++ b/package.use/plasma-meta @@ -1 +1 @@ -kde-plasma/plasma-meta -browser-integration -crypt -wallpapers +kde-plasma/plasma-meta gtk -browser-integration -crypt -wallpapers diff --git a/package.use/polkit-qt b/package.use/polkit-qt new file mode 100644 index 0000000..865316c --- /dev/null +++ b/package.use/polkit-qt @@ -0,0 +1 @@ +sys-auth/polkit-qt qt5 diff --git a/package.use/qtbase b/package.use/qtbase index 98ded9f..6941c14 100644 --- a/package.use/qtbase +++ b/package.use/qtbase @@ -1 +1 @@ -dev-qt/qtbase -dbus -libinpput -libproxy +dev-qt/qtbase cups diff --git a/patches/sys-kernel/gentoo-sources/0001-add-sysctl-to-disallow-unprivileged-CLONE_NEWUSER-by.patch b/patches/sys-kernel/gentoo-sources/0001-add-sysctl-to-disallow-unprivileged-CLONE_NEWUSER-by.patch new file mode 100644 index 0000000..30a77c7 --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0001-add-sysctl-to-disallow-unprivileged-CLONE_NEWUSER-by.patch @@ -0,0 +1,151 @@ +From d50977b164e708bf523a35ef53315355528c3ca6 Mon Sep 17 00:00:00 2001 +From: "Jan Alexander Steffens (heftig)" +Date: Mon, 16 Sep 2019 04:53:20 +0200 +Subject: [PATCH] ZEN: Add sysctl and CONFIG to disallow unprivileged + CLONE_NEWUSER + +Our default behavior continues to match the vanilla kernel. +--- + include/linux/user_namespace.h | 4 ++++ + init/Kconfig | 16 ++++++++++++++++ + kernel/fork.c | 14 ++++++++++++++ + kernel/sysctl.c | 12 ++++++++++++ + kernel/user_namespace.c | 7 +++++++ + 5 files changed, 53 insertions(+) + +diff --git a/include/linux/user_namespace.h b/include/linux/user_namespace.h +index 45f09bec02c485..87b20e2ee27445 100644 +--- a/include/linux/user_namespace.h ++++ b/include/linux/user_namespace.h +@@ -148,6 +148,8 @@ static inline void set_userns_rlimit_max(struct user_namespace *ns, + + #ifdef CONFIG_USER_NS + ++extern int unprivileged_userns_clone; ++ + static inline struct user_namespace *get_user_ns(struct user_namespace *ns) + { + if (ns) +@@ -181,6 +183,8 @@ extern bool current_in_userns(const struct user_namespace *target_ns); + struct ns_common *ns_get_owner(struct ns_common *ns); + #else + ++#define unprivileged_userns_clone 0 ++ + static inline struct user_namespace *get_user_ns(struct user_namespace *ns) + { + return &init_user_ns; +diff --git a/init/Kconfig b/init/Kconfig +index 94125d3b6893c7..9f7139b536f638 100644 +--- a/init/Kconfig ++++ b/init/Kconfig +@@ -1247,6 +1247,22 @@ config USER_NS + + If unsure, say N. + ++config USER_NS_UNPRIVILEGED ++ bool "Allow unprivileged users to create namespaces" ++ default y ++ depends on USER_NS ++ help ++ When disabled, unprivileged users will not be able to create ++ new namespaces. Allowing users to create their own namespaces ++ has been part of several recent local privilege escalation ++ exploits, so if you need user namespaces but are ++ paranoid^Wsecurity-conscious you want to disable this. ++ ++ This setting can be overridden at runtime via the ++ kernel.unprivileged_userns_clone sysctl. ++ ++ If unsure, say Y. ++ + config PID_NS + bool "PID Namespaces" + default y +diff --git a/kernel/fork.c b/kernel/fork.c +index 08969f5aa38d59..ff601cb7a1fae0 100644 +--- a/kernel/fork.c ++++ b/kernel/fork.c +@@ -98,6 +98,10 @@ + #include + #include + ++#ifdef CONFIG_USER_NS ++#include ++#endif ++ + #include + #include + #include +@@ -2008,6 +2012,10 @@ static __latent_entropy struct task_struct *copy_process( + if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) + return ERR_PTR(-EINVAL); + ++ if ((clone_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) ++ if (!capable(CAP_SYS_ADMIN)) ++ return ERR_PTR(-EPERM); ++ + /* + * Thread groups must share signals as well, and detached threads + * can only be started up within the thread group. +@@ -3166,6 +3174,12 @@ int ksys_unshare(unsigned long unshare_flags) + if (unshare_flags & CLONE_NEWNS) + unshare_flags |= CLONE_FS; + ++ if ((unshare_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) { ++ err = -EPERM; ++ if (!capable(CAP_SYS_ADMIN)) ++ goto bad_unshare_out; ++ } ++ + err = check_unshare_flags(unshare_flags); + if (err) + goto bad_unshare_out; +diff --git a/kernel/sysctl.c b/kernel/sysctl.c +index c6d9dec11b749d..9a4514ad481b21 100644 +--- a/kernel/sysctl.c ++++ b/kernel/sysctl.c +@@ -81,6 +81,9 @@ + #ifdef CONFIG_RT_MUTEXES + #include + #endif ++#ifdef CONFIG_USER_NS ++#include ++#endif + + /* shared constants to be used in various sysctls */ + const int sysctl_vals[] = { 0, 1, 2, 3, 4, 100, 200, 1000, 3000, INT_MAX, 65535, -1 }; +@@ -1659,6 +1662,15 @@ static struct ctl_table kern_table[] = { + .mode = 0644, + .proc_handler = proc_dointvec, + }, ++#ifdef CONFIG_USER_NS ++ { ++ .procname = "unprivileged_userns_clone", ++ .data = &unprivileged_userns_clone, ++ .maxlen = sizeof(int), ++ .mode = 0644, ++ .proc_handler = proc_dointvec, ++ }, ++#endif + #ifdef CONFIG_PROC_SYSCTL + { + .procname = "tainted", +diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c +index 54211dbd516c57..16ca0c1516298d 100644 +--- a/kernel/user_namespace.c ++++ b/kernel/user_namespace.c +@@ -22,6 +22,13 @@ + #include + #include + ++/* sysctl */ ++#ifdef CONFIG_USER_NS_UNPRIVILEGED ++int unprivileged_userns_clone = 1; ++#else ++int unprivileged_userns_clone; ++#endif ++ + static struct kmem_cache *user_ns_cachep __ro_after_init; + static DEFINE_MUTEX(userns_state_mutex); + diff --git a/patches/sys-kernel/gentoo-sources/0002-bbr3.patch b/patches/sys-kernel/gentoo-sources/0002-bbr3.patch new file mode 100644 index 0000000..f9453a9 --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0002-bbr3.patch @@ -0,0 +1,3386 @@ +From 3e76598e1f8e7468aa7ceaee2e1ebc4683c03ed5 Mon Sep 17 00:00:00 2001 +From: Peter Jung +Date: Mon, 7 Oct 2024 11:53:09 +0200 +Subject: [PATCH 2/8] bbr3 + +Signed-off-by: Peter Jung +--- + include/linux/tcp.h | 4 +- + include/net/inet_connection_sock.h | 4 +- + include/net/tcp.h | 72 +- + include/uapi/linux/inet_diag.h | 23 + + include/uapi/linux/rtnetlink.h | 4 +- + include/uapi/linux/tcp.h | 1 + + net/ipv4/Kconfig | 21 +- + net/ipv4/bpf_tcp_ca.c | 9 +- + net/ipv4/tcp.c | 3 + + net/ipv4/tcp_bbr.c | 2230 +++++++++++++++++++++------- + net/ipv4/tcp_cong.c | 1 + + net/ipv4/tcp_input.c | 40 +- + net/ipv4/tcp_minisocks.c | 2 + + net/ipv4/tcp_output.c | 48 +- + net/ipv4/tcp_rate.c | 30 +- + net/ipv4/tcp_timer.c | 1 + + 16 files changed, 1940 insertions(+), 553 deletions(-) + +diff --git a/include/linux/tcp.h b/include/linux/tcp.h +index 6a5e08b937b3..27aab715490e 100644 +--- a/include/linux/tcp.h ++++ b/include/linux/tcp.h +@@ -369,7 +369,9 @@ struct tcp_sock { + u8 compressed_ack; + u8 dup_ack_counter:2, + tlp_retrans:1, /* TLP is a retransmission */ +- unused:5; ++ fast_ack_mode:2, /* which fast ack mode ? */ ++ tlp_orig_data_app_limited:1, /* app-limited before TLP rtx? */ ++ unused:2; + u8 thin_lto : 1,/* Use linear timeouts for thin streams */ + fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */ + fastopen_no_cookie:1, /* Allow send/recv SYN+data without a cookie */ +diff --git a/include/net/inet_connection_sock.h b/include/net/inet_connection_sock.h +index c0deaafebfdc..d53f042d936e 100644 +--- a/include/net/inet_connection_sock.h ++++ b/include/net/inet_connection_sock.h +@@ -137,8 +137,8 @@ struct inet_connection_sock { + u32 icsk_probes_tstamp; + u32 icsk_user_timeout; + +- u64 icsk_ca_priv[104 / sizeof(u64)]; +-#define ICSK_CA_PRIV_SIZE sizeof_field(struct inet_connection_sock, icsk_ca_priv) ++#define ICSK_CA_PRIV_SIZE (144) ++ u64 icsk_ca_priv[ICSK_CA_PRIV_SIZE / sizeof(u64)]; + }; + + #define ICSK_TIME_RETRANS 1 /* Retransmit timer */ +diff --git a/include/net/tcp.h b/include/net/tcp.h +index d1948d357dad..7d99f0bec5f2 100644 +--- a/include/net/tcp.h ++++ b/include/net/tcp.h +@@ -375,6 +375,8 @@ static inline void tcp_dec_quickack_mode(struct sock *sk) + #define TCP_ECN_QUEUE_CWR 2 + #define TCP_ECN_DEMAND_CWR 4 + #define TCP_ECN_SEEN 8 ++#define TCP_ECN_LOW 16 ++#define TCP_ECN_ECT_PERMANENT 32 + + enum tcp_tw_status { + TCP_TW_SUCCESS = 0, +@@ -779,6 +781,15 @@ static inline void tcp_fast_path_check(struct sock *sk) + + u32 tcp_delack_max(const struct sock *sk); + ++static inline void tcp_set_ecn_low_from_dst(struct sock *sk, ++ const struct dst_entry *dst) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ ++ if (dst_feature(dst, RTAX_FEATURE_ECN_LOW)) ++ tp->ecn_flags |= TCP_ECN_LOW; ++} ++ + /* Compute the actual rto_min value */ + static inline u32 tcp_rto_min(const struct sock *sk) + { +@@ -884,6 +895,11 @@ static inline u32 tcp_stamp_us_delta(u64 t1, u64 t0) + return max_t(s64, t1 - t0, 0); + } + ++static inline u32 tcp_stamp32_us_delta(u32 t1, u32 t0) ++{ ++ return max_t(s32, t1 - t0, 0); ++} ++ + /* provide the departure time in us unit */ + static inline u64 tcp_skb_timestamp_us(const struct sk_buff *skb) + { +@@ -973,9 +989,14 @@ struct tcp_skb_cb { + /* pkts S/ACKed so far upon tx of skb, incl retrans: */ + __u32 delivered; + /* start of send pipeline phase */ +- u64 first_tx_mstamp; ++ u32 first_tx_mstamp; + /* when we reached the "delivered" count */ +- u64 delivered_mstamp; ++ u32 delivered_mstamp; ++#define TCPCB_IN_FLIGHT_BITS 20 ++#define TCPCB_IN_FLIGHT_MAX ((1U << TCPCB_IN_FLIGHT_BITS) - 1) ++ u32 in_flight:20, /* packets in flight at transmit */ ++ unused2:12; ++ u32 lost; /* packets lost so far upon tx of skb */ + } tx; /* only used for outgoing skbs */ + union { + struct inet_skb_parm h4; +@@ -1088,6 +1109,7 @@ enum tcp_ca_event { + CA_EVENT_LOSS, /* loss timeout */ + CA_EVENT_ECN_NO_CE, /* ECT set, but not CE marked */ + CA_EVENT_ECN_IS_CE, /* received CE marked IP packet */ ++ CA_EVENT_TLP_RECOVERY, /* a lost segment was repaired by TLP probe */ + }; + + /* Information about inbound ACK, passed to cong_ops->in_ack_event() */ +@@ -1110,7 +1132,11 @@ enum tcp_ca_ack_event_flags { + #define TCP_CONG_NON_RESTRICTED 0x1 + /* Requires ECN/ECT set on all packets */ + #define TCP_CONG_NEEDS_ECN 0x2 +-#define TCP_CONG_MASK (TCP_CONG_NON_RESTRICTED | TCP_CONG_NEEDS_ECN) ++/* Wants notification of CE events (CA_EVENT_ECN_IS_CE, CA_EVENT_ECN_NO_CE). */ ++#define TCP_CONG_WANTS_CE_EVENTS 0x4 ++#define TCP_CONG_MASK (TCP_CONG_NON_RESTRICTED | \ ++ TCP_CONG_NEEDS_ECN | \ ++ TCP_CONG_WANTS_CE_EVENTS) + + union tcp_cc_info; + +@@ -1130,10 +1156,13 @@ struct ack_sample { + */ + struct rate_sample { + u64 prior_mstamp; /* starting timestamp for interval */ ++ u32 prior_lost; /* tp->lost at "prior_mstamp" */ + u32 prior_delivered; /* tp->delivered at "prior_mstamp" */ + u32 prior_delivered_ce;/* tp->delivered_ce at "prior_mstamp" */ ++ u32 tx_in_flight; /* packets in flight at starting timestamp */ ++ s32 lost; /* number of packets lost over interval */ + s32 delivered; /* number of packets delivered over interval */ +- s32 delivered_ce; /* number of packets delivered w/ CE marks*/ ++ s32 delivered_ce; /* packets delivered w/ CE mark over interval */ + long interval_us; /* time for tp->delivered to incr "delivered" */ + u32 snd_interval_us; /* snd interval for delivered packets */ + u32 rcv_interval_us; /* rcv interval for delivered packets */ +@@ -1144,7 +1173,9 @@ struct rate_sample { + u32 last_end_seq; /* end_seq of most recently ACKed packet */ + bool is_app_limited; /* is sample from packet with bubble in pipe? */ + bool is_retrans; /* is sample from retransmission? */ ++ bool is_acking_tlp_retrans_seq; /* ACKed a TLP retransmit sequence? */ + bool is_ack_delayed; /* is this (likely) a delayed ACK? */ ++ bool is_ece; /* did this ACK have ECN marked? */ + }; + + struct tcp_congestion_ops { +@@ -1168,8 +1199,11 @@ struct tcp_congestion_ops { + /* hook for packet ack accounting (optional) */ + void (*pkts_acked)(struct sock *sk, const struct ack_sample *sample); + +- /* override sysctl_tcp_min_tso_segs */ +- u32 (*min_tso_segs)(struct sock *sk); ++ /* pick target number of segments per TSO/GSO skb (optional): */ ++ u32 (*tso_segs)(struct sock *sk, unsigned int mss_now); ++ ++ /* react to a specific lost skb (optional) */ ++ void (*skb_marked_lost)(struct sock *sk, const struct sk_buff *skb); + + /* call when packets are delivered to update cwnd and pacing rate, + * after all the ca_state processing. (optional) +@@ -1235,6 +1269,14 @@ static inline char *tcp_ca_get_name_by_key(u32 key, char *buffer) + } + #endif + ++static inline bool tcp_ca_wants_ce_events(const struct sock *sk) ++{ ++ const struct inet_connection_sock *icsk = inet_csk(sk); ++ ++ return icsk->icsk_ca_ops->flags & (TCP_CONG_NEEDS_ECN | ++ TCP_CONG_WANTS_CE_EVENTS); ++} ++ + static inline bool tcp_ca_needs_ecn(const struct sock *sk) + { + const struct inet_connection_sock *icsk = inet_csk(sk); +@@ -1254,6 +1296,7 @@ static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) + void tcp_set_ca_state(struct sock *sk, const u8 ca_state); + + /* From tcp_rate.c */ ++void tcp_set_tx_in_flight(struct sock *sk, struct sk_buff *skb); + void tcp_rate_skb_sent(struct sock *sk, struct sk_buff *skb); + void tcp_rate_skb_delivered(struct sock *sk, struct sk_buff *skb, + struct rate_sample *rs); +@@ -1266,6 +1309,21 @@ static inline bool tcp_skb_sent_after(u64 t1, u64 t2, u32 seq1, u32 seq2) + return t1 > t2 || (t1 == t2 && after(seq1, seq2)); + } + ++/* If a retransmit failed due to local qdisc congestion or other local issues, ++ * then we may have called tcp_set_skb_tso_segs() to increase the number of ++ * segments in the skb without increasing the tx.in_flight. In all other cases, ++ * the tx.in_flight should be at least as big as the pcount of the sk_buff. We ++ * do not have the state to know whether a retransmit failed due to local qdisc ++ * congestion or other local issues, so to avoid spurious warnings we consider ++ * that any skb marked lost may have suffered that fate. ++ */ ++static inline bool tcp_skb_tx_in_flight_is_suspicious(u32 skb_pcount, ++ u32 skb_sacked_flags, ++ u32 tx_in_flight) ++{ ++ return (skb_pcount > tx_in_flight) && !(skb_sacked_flags & TCPCB_LOST); ++} ++ + /* These functions determine how the current flow behaves in respect of SACK + * handling. SACK is negotiated with the peer, and therefore it can vary + * between different flows. +@@ -2417,7 +2475,7 @@ struct tcp_plb_state { + u8 consec_cong_rounds:5, /* consecutive congested rounds */ + unused:3; + u32 pause_until; /* jiffies32 when PLB can resume rerouting */ +-}; ++} __attribute__ ((__packed__)); + + static inline void tcp_plb_init(const struct sock *sk, + struct tcp_plb_state *plb) +diff --git a/include/uapi/linux/inet_diag.h b/include/uapi/linux/inet_diag.h +index 86bb2e8b17c9..9d9a3eb2ce9b 100644 +--- a/include/uapi/linux/inet_diag.h ++++ b/include/uapi/linux/inet_diag.h +@@ -229,6 +229,29 @@ struct tcp_bbr_info { + __u32 bbr_min_rtt; /* min-filtered RTT in uSec */ + __u32 bbr_pacing_gain; /* pacing gain shifted left 8 bits */ + __u32 bbr_cwnd_gain; /* cwnd gain shifted left 8 bits */ ++ __u32 bbr_bw_hi_lsb; /* lower 32 bits of bw_hi */ ++ __u32 bbr_bw_hi_msb; /* upper 32 bits of bw_hi */ ++ __u32 bbr_bw_lo_lsb; /* lower 32 bits of bw_lo */ ++ __u32 bbr_bw_lo_msb; /* upper 32 bits of bw_lo */ ++ __u8 bbr_mode; /* current bbr_mode in state machine */ ++ __u8 bbr_phase; /* current state machine phase */ ++ __u8 unused1; /* alignment padding; not used yet */ ++ __u8 bbr_version; /* BBR algorithm version */ ++ __u32 bbr_inflight_lo; /* lower short-term data volume bound */ ++ __u32 bbr_inflight_hi; /* higher long-term data volume bound */ ++ __u32 bbr_extra_acked; /* max excess packets ACKed in epoch */ ++}; ++ ++/* TCP BBR congestion control bbr_phase as reported in netlink/ss stats. */ ++enum tcp_bbr_phase { ++ BBR_PHASE_INVALID = 0, ++ BBR_PHASE_STARTUP = 1, ++ BBR_PHASE_DRAIN = 2, ++ BBR_PHASE_PROBE_RTT = 3, ++ BBR_PHASE_PROBE_BW_UP = 4, ++ BBR_PHASE_PROBE_BW_DOWN = 5, ++ BBR_PHASE_PROBE_BW_CRUISE = 6, ++ BBR_PHASE_PROBE_BW_REFILL = 7, + }; + + union tcp_cc_info { +diff --git a/include/uapi/linux/rtnetlink.h b/include/uapi/linux/rtnetlink.h +index 3b687d20c9ed..a7c30c243b54 100644 +--- a/include/uapi/linux/rtnetlink.h ++++ b/include/uapi/linux/rtnetlink.h +@@ -507,12 +507,14 @@ enum { + #define RTAX_FEATURE_TIMESTAMP (1 << 2) /* unused */ + #define RTAX_FEATURE_ALLFRAG (1 << 3) /* unused */ + #define RTAX_FEATURE_TCP_USEC_TS (1 << 4) ++#define RTAX_FEATURE_ECN_LOW (1 << 5) + + #define RTAX_FEATURE_MASK (RTAX_FEATURE_ECN | \ + RTAX_FEATURE_SACK | \ + RTAX_FEATURE_TIMESTAMP | \ + RTAX_FEATURE_ALLFRAG | \ +- RTAX_FEATURE_TCP_USEC_TS) ++ RTAX_FEATURE_TCP_USEC_TS | \ ++ RTAX_FEATURE_ECN_LOW) + + struct rta_session { + __u8 proto; +diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h +index dbf896f3146c..4702cd2f1ffc 100644 +--- a/include/uapi/linux/tcp.h ++++ b/include/uapi/linux/tcp.h +@@ -178,6 +178,7 @@ enum tcp_fastopen_client_fail { + #define TCPI_OPT_ECN_SEEN 16 /* we received at least one packet with ECT */ + #define TCPI_OPT_SYN_DATA 32 /* SYN-ACK acked data in SYN sent or rcvd */ + #define TCPI_OPT_USEC_TS 64 /* usec timestamps */ ++#define TCPI_OPT_ECN_LOW 128 /* Low-latency ECN configured at init */ + + /* + * Sender's congestion state indicating normal or abnormal situations +diff --git a/net/ipv4/Kconfig b/net/ipv4/Kconfig +index 6d2c97f8e9ef..ddc116ef22cb 100644 +--- a/net/ipv4/Kconfig ++++ b/net/ipv4/Kconfig +@@ -669,15 +669,18 @@ config TCP_CONG_BBR + default n + help + +- BBR (Bottleneck Bandwidth and RTT) TCP congestion control aims to +- maximize network utilization and minimize queues. It builds an explicit +- model of the bottleneck delivery rate and path round-trip propagation +- delay. It tolerates packet loss and delay unrelated to congestion. It +- can operate over LAN, WAN, cellular, wifi, or cable modem links. It can +- coexist with flows that use loss-based congestion control, and can +- operate with shallow buffers, deep buffers, bufferbloat, policers, or +- AQM schemes that do not provide a delay signal. It requires the fq +- ("Fair Queue") pacing packet scheduler. ++ BBR (Bottleneck Bandwidth and RTT) TCP congestion control is a ++ model-based congestion control algorithm that aims to maximize ++ network utilization, keep queues and retransmit rates low, and to be ++ able to coexist with Reno/CUBIC in common scenarios. It builds an ++ explicit model of the network path. It tolerates a targeted degree ++ of random packet loss and delay. It can operate over LAN, WAN, ++ cellular, wifi, or cable modem links, and can use shallow-threshold ++ ECN signals. It can coexist to some degree with flows that use ++ loss-based congestion control, and can operate with shallow buffers, ++ deep buffers, bufferbloat, policers, or AQM schemes that do not ++ provide a delay signal. It requires pacing, using either TCP internal ++ pacing or the fq ("Fair Queue") pacing packet scheduler. + + choice + prompt "Default TCP congestion control" +diff --git a/net/ipv4/bpf_tcp_ca.c b/net/ipv4/bpf_tcp_ca.c +index 554804774628..2279e6e7bc9c 100644 +--- a/net/ipv4/bpf_tcp_ca.c ++++ b/net/ipv4/bpf_tcp_ca.c +@@ -280,11 +280,15 @@ static void bpf_tcp_ca_pkts_acked(struct sock *sk, const struct ack_sample *samp + { + } + +-static u32 bpf_tcp_ca_min_tso_segs(struct sock *sk) ++static u32 bpf_tcp_ca_tso_segs(struct sock *sk, unsigned int mss_now) + { + return 0; + } + ++static void bpf_tcp_ca_skb_marked_lost(struct sock *sk, const struct sk_buff *skb) ++{ ++} ++ + static void bpf_tcp_ca_cong_control(struct sock *sk, u32 ack, int flag, + const struct rate_sample *rs) + { +@@ -315,7 +319,8 @@ static struct tcp_congestion_ops __bpf_ops_tcp_congestion_ops = { + .cwnd_event = bpf_tcp_ca_cwnd_event, + .in_ack_event = bpf_tcp_ca_in_ack_event, + .pkts_acked = bpf_tcp_ca_pkts_acked, +- .min_tso_segs = bpf_tcp_ca_min_tso_segs, ++ .tso_segs = bpf_tcp_ca_tso_segs, ++ .skb_marked_lost = bpf_tcp_ca_skb_marked_lost, + .cong_control = bpf_tcp_ca_cong_control, + .undo_cwnd = bpf_tcp_ca_undo_cwnd, + .sndbuf_expand = bpf_tcp_ca_sndbuf_expand, +diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c +index 4f77bd862e95..fd3a5551eda7 100644 +--- a/net/ipv4/tcp.c ++++ b/net/ipv4/tcp.c +@@ -3384,6 +3384,7 @@ int tcp_disconnect(struct sock *sk, int flags) + tp->rx_opt.dsack = 0; + tp->rx_opt.num_sacks = 0; + tp->rcv_ooopack = 0; ++ tp->fast_ack_mode = 0; + + + /* Clean up fastopen related fields */ +@@ -4110,6 +4111,8 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info) + info->tcpi_options |= TCPI_OPT_ECN; + if (tp->ecn_flags & TCP_ECN_SEEN) + info->tcpi_options |= TCPI_OPT_ECN_SEEN; ++ if (tp->ecn_flags & TCP_ECN_LOW) ++ info->tcpi_options |= TCPI_OPT_ECN_LOW; + if (tp->syn_data_acked) + info->tcpi_options |= TCPI_OPT_SYN_DATA; + if (tp->tcp_usec_ts) +diff --git a/net/ipv4/tcp_bbr.c b/net/ipv4/tcp_bbr.c +index 760941e55153..a180fa648d5e 100644 +--- a/net/ipv4/tcp_bbr.c ++++ b/net/ipv4/tcp_bbr.c +@@ -1,18 +1,19 @@ +-/* Bottleneck Bandwidth and RTT (BBR) congestion control ++/* BBR (Bottleneck Bandwidth and RTT) congestion control + * +- * BBR congestion control computes the sending rate based on the delivery +- * rate (throughput) estimated from ACKs. In a nutshell: ++ * BBR is a model-based congestion control algorithm that aims for low queues, ++ * low loss, and (bounded) Reno/CUBIC coexistence. To maintain a model of the ++ * network path, it uses measurements of bandwidth and RTT, as well as (if they ++ * occur) packet loss and/or shallow-threshold ECN signals. Note that although ++ * it can use ECN or loss signals explicitly, it does not require either; it ++ * can bound its in-flight data based on its estimate of the BDP. + * +- * On each ACK, update our model of the network path: +- * bottleneck_bandwidth = windowed_max(delivered / elapsed, 10 round trips) +- * min_rtt = windowed_min(rtt, 10 seconds) +- * pacing_rate = pacing_gain * bottleneck_bandwidth +- * cwnd = max(cwnd_gain * bottleneck_bandwidth * min_rtt, 4) +- * +- * The core algorithm does not react directly to packet losses or delays, +- * although BBR may adjust the size of next send per ACK when loss is +- * observed, or adjust the sending rate if it estimates there is a +- * traffic policer, in order to keep the drop rate reasonable. ++ * The model has both higher and lower bounds for the operating range: ++ * lo: bw_lo, inflight_lo: conservative short-term lower bound ++ * hi: bw_hi, inflight_hi: robust long-term upper bound ++ * The bandwidth-probing time scale is (a) extended dynamically based on ++ * estimated BDP to improve coexistence with Reno/CUBIC; (b) bounded by ++ * an interactive wall-clock time-scale to be more scalable and responsive ++ * than Reno and CUBIC. + * + * Here is a state transition diagram for BBR: + * +@@ -65,6 +66,13 @@ + #include + #include + ++#include ++#include "tcp_dctcp.h" ++ ++#define BBR_VERSION 3 ++ ++#define bbr_param(sk,name) (bbr_ ## name) ++ + /* Scale factor for rate in pkt/uSec unit to avoid truncation in bandwidth + * estimation. The rate unit ~= (1500 bytes / 1 usec / 2^24) ~= 715 bps. + * This handles bandwidths from 0.06pps (715bps) to 256Mpps (3Tbps) in a u32. +@@ -85,36 +93,41 @@ enum bbr_mode { + BBR_PROBE_RTT, /* cut inflight to min to probe min_rtt */ + }; + ++/* How does the incoming ACK stream relate to our bandwidth probing? */ ++enum bbr_ack_phase { ++ BBR_ACKS_INIT, /* not probing; not getting probe feedback */ ++ BBR_ACKS_REFILLING, /* sending at est. bw to fill pipe */ ++ BBR_ACKS_PROBE_STARTING, /* inflight rising to probe bw */ ++ BBR_ACKS_PROBE_FEEDBACK, /* getting feedback from bw probing */ ++ BBR_ACKS_PROBE_STOPPING, /* stopped probing; still getting feedback */ ++}; ++ + /* BBR congestion control block */ + struct bbr { + u32 min_rtt_us; /* min RTT in min_rtt_win_sec window */ + u32 min_rtt_stamp; /* timestamp of min_rtt_us */ + u32 probe_rtt_done_stamp; /* end time for BBR_PROBE_RTT mode */ +- struct minmax bw; /* Max recent delivery rate in pkts/uS << 24 */ +- u32 rtt_cnt; /* count of packet-timed rounds elapsed */ ++ u32 probe_rtt_min_us; /* min RTT in probe_rtt_win_ms win */ ++ u32 probe_rtt_min_stamp; /* timestamp of probe_rtt_min_us*/ + u32 next_rtt_delivered; /* scb->tx.delivered at end of round */ + u64 cycle_mstamp; /* time of this cycle phase start */ +- u32 mode:3, /* current bbr_mode in state machine */ ++ u32 mode:2, /* current bbr_mode in state machine */ + prev_ca_state:3, /* CA state on previous ACK */ +- packet_conservation:1, /* use packet conservation? */ + round_start:1, /* start of packet-timed tx->ack round? */ ++ ce_state:1, /* If most recent data has CE bit set */ ++ bw_probe_up_rounds:5, /* cwnd-limited rounds in PROBE_UP */ ++ try_fast_path:1, /* can we take fast path? */ + idle_restart:1, /* restarting after idle? */ + probe_rtt_round_done:1, /* a BBR_PROBE_RTT round at 4 pkts? */ +- unused:13, +- lt_is_sampling:1, /* taking long-term ("LT") samples now? */ +- lt_rtt_cnt:7, /* round trips in long-term interval */ +- lt_use_bw:1; /* use lt_bw as our bw estimate? */ +- u32 lt_bw; /* LT est delivery rate in pkts/uS << 24 */ +- u32 lt_last_delivered; /* LT intvl start: tp->delivered */ +- u32 lt_last_stamp; /* LT intvl start: tp->delivered_mstamp */ +- u32 lt_last_lost; /* LT intvl start: tp->lost */ ++ init_cwnd:7, /* initial cwnd */ ++ unused_1:10; + u32 pacing_gain:10, /* current gain for setting pacing rate */ + cwnd_gain:10, /* current gain for setting cwnd */ + full_bw_reached:1, /* reached full bw in Startup? */ + full_bw_cnt:2, /* number of rounds without large bw gains */ +- cycle_idx:3, /* current index in pacing_gain cycle array */ ++ cycle_idx:2, /* current index in pacing_gain cycle array */ + has_seen_rtt:1, /* have we seen an RTT sample yet? */ +- unused_b:5; ++ unused_2:6; + u32 prior_cwnd; /* prior cwnd upon entering loss recovery */ + u32 full_bw; /* recent bw, to estimate if pipe is full */ + +@@ -124,19 +137,67 @@ struct bbr { + u32 ack_epoch_acked:20, /* packets (S)ACKed in sampling epoch */ + extra_acked_win_rtts:5, /* age of extra_acked, in round trips */ + extra_acked_win_idx:1, /* current index in extra_acked array */ +- unused_c:6; ++ /* BBR v3 state: */ ++ full_bw_now:1, /* recently reached full bw plateau? */ ++ startup_ecn_rounds:2, /* consecutive hi ECN STARTUP rounds */ ++ loss_in_cycle:1, /* packet loss in this cycle? */ ++ ecn_in_cycle:1, /* ECN in this cycle? */ ++ unused_3:1; ++ u32 loss_round_delivered; /* scb->tx.delivered ending loss round */ ++ u32 undo_bw_lo; /* bw_lo before latest losses */ ++ u32 undo_inflight_lo; /* inflight_lo before latest losses */ ++ u32 undo_inflight_hi; /* inflight_hi before latest losses */ ++ u32 bw_latest; /* max delivered bw in last round trip */ ++ u32 bw_lo; /* lower bound on sending bandwidth */ ++ u32 bw_hi[2]; /* max recent measured bw sample */ ++ u32 inflight_latest; /* max delivered data in last round trip */ ++ u32 inflight_lo; /* lower bound of inflight data range */ ++ u32 inflight_hi; /* upper bound of inflight data range */ ++ u32 bw_probe_up_cnt; /* packets delivered per inflight_hi incr */ ++ u32 bw_probe_up_acks; /* packets (S)ACKed since inflight_hi incr */ ++ u32 probe_wait_us; /* PROBE_DOWN until next clock-driven probe */ ++ u32 prior_rcv_nxt; /* tp->rcv_nxt when CE state last changed */ ++ u32 ecn_eligible:1, /* sender can use ECN (RTT, handshake)? */ ++ ecn_alpha:9, /* EWMA delivered_ce/delivered; 0..256 */ ++ bw_probe_samples:1, /* rate samples reflect bw probing? */ ++ prev_probe_too_high:1, /* did last PROBE_UP go too high? */ ++ stopped_risky_probe:1, /* last PROBE_UP stopped due to risk? */ ++ rounds_since_probe:8, /* packet-timed rounds since probed bw */ ++ loss_round_start:1, /* loss_round_delivered round trip? */ ++ loss_in_round:1, /* loss marked in this round trip? */ ++ ecn_in_round:1, /* ECN marked in this round trip? */ ++ ack_phase:3, /* bbr_ack_phase: meaning of ACKs */ ++ loss_events_in_round:4,/* losses in STARTUP round */ ++ initialized:1; /* has bbr_init() been called? */ ++ u32 alpha_last_delivered; /* tp->delivered at alpha update */ ++ u32 alpha_last_delivered_ce; /* tp->delivered_ce at alpha update */ ++ ++ u8 unused_4; /* to preserve alignment */ ++ struct tcp_plb_state plb; + }; + +-#define CYCLE_LEN 8 /* number of phases in a pacing gain cycle */ ++struct bbr_context { ++ u32 sample_bw; ++}; + +-/* Window length of bw filter (in rounds): */ +-static const int bbr_bw_rtts = CYCLE_LEN + 2; + /* Window length of min_rtt filter (in sec): */ + static const u32 bbr_min_rtt_win_sec = 10; + /* Minimum time (in ms) spent at bbr_cwnd_min_target in BBR_PROBE_RTT mode: */ + static const u32 bbr_probe_rtt_mode_ms = 200; +-/* Skip TSO below the following bandwidth (bits/sec): */ +-static const int bbr_min_tso_rate = 1200000; ++/* Window length of probe_rtt_min_us filter (in ms), and consequently the ++ * typical interval between PROBE_RTT mode entries. The default is 5000ms. ++ * Note that bbr_probe_rtt_win_ms must be <= bbr_min_rtt_win_sec * MSEC_PER_SEC ++ */ ++static const u32 bbr_probe_rtt_win_ms = 5000; ++/* Proportion of cwnd to estimated BDP in PROBE_RTT, in units of BBR_UNIT: */ ++static const u32 bbr_probe_rtt_cwnd_gain = BBR_UNIT * 1 / 2; ++ ++/* Use min_rtt to help adapt TSO burst size, with smaller min_rtt resulting ++ * in bigger TSO bursts. We cut the RTT-based allowance in half ++ * for every 2^9 usec (aka 512 us) of RTT, so that the RTT-based allowance ++ * is below 1500 bytes after 6 * ~500 usec = 3ms. ++ */ ++static const u32 bbr_tso_rtt_shift = 9; + + /* Pace at ~1% below estimated bw, on average, to reduce queue at bottleneck. + * In order to help drive the network toward lower queues and low latency while +@@ -146,13 +207,15 @@ static const int bbr_min_tso_rate = 1200000; + */ + static const int bbr_pacing_margin_percent = 1; + +-/* We use a high_gain value of 2/ln(2) because it's the smallest pacing gain ++/* We use a startup_pacing_gain of 4*ln(2) because it's the smallest value + * that will allow a smoothly increasing pacing rate that will double each RTT + * and send the same number of packets per RTT that an un-paced, slow-starting + * Reno or CUBIC flow would: + */ +-static const int bbr_high_gain = BBR_UNIT * 2885 / 1000 + 1; +-/* The pacing gain of 1/high_gain in BBR_DRAIN is calculated to typically drain ++static const int bbr_startup_pacing_gain = BBR_UNIT * 277 / 100 + 1; ++/* The gain for deriving startup cwnd: */ ++static const int bbr_startup_cwnd_gain = BBR_UNIT * 2; ++/* The pacing gain in BBR_DRAIN is calculated to typically drain + * the queue created in BBR_STARTUP in a single round: + */ + static const int bbr_drain_gain = BBR_UNIT * 1000 / 2885; +@@ -160,13 +223,17 @@ static const int bbr_drain_gain = BBR_UNIT * 1000 / 2885; + static const int bbr_cwnd_gain = BBR_UNIT * 2; + /* The pacing_gain values for the PROBE_BW gain cycle, to discover/share bw: */ + static const int bbr_pacing_gain[] = { +- BBR_UNIT * 5 / 4, /* probe for more available bw */ +- BBR_UNIT * 3 / 4, /* drain queue and/or yield bw to other flows */ +- BBR_UNIT, BBR_UNIT, BBR_UNIT, /* cruise at 1.0*bw to utilize pipe, */ +- BBR_UNIT, BBR_UNIT, BBR_UNIT /* without creating excess queue... */ ++ BBR_UNIT * 5 / 4, /* UP: probe for more available bw */ ++ BBR_UNIT * 91 / 100, /* DOWN: drain queue and/or yield bw */ ++ BBR_UNIT, /* CRUISE: try to use pipe w/ some headroom */ ++ BBR_UNIT, /* REFILL: refill pipe to estimated 100% */ ++}; ++enum bbr_pacing_gain_phase { ++ BBR_BW_PROBE_UP = 0, /* push up inflight to probe for bw/vol */ ++ BBR_BW_PROBE_DOWN = 1, /* drain excess inflight from the queue */ ++ BBR_BW_PROBE_CRUISE = 2, /* use pipe, w/ headroom in queue/pipe */ ++ BBR_BW_PROBE_REFILL = 3, /* v2: refill the pipe again to 100% */ + }; +-/* Randomize the starting gain cycling phase over N phases: */ +-static const u32 bbr_cycle_rand = 7; + + /* Try to keep at least this many packets in flight, if things go smoothly. For + * smooth functioning, a sliding window protocol ACKing every other packet +@@ -174,24 +241,12 @@ static const u32 bbr_cycle_rand = 7; + */ + static const u32 bbr_cwnd_min_target = 4; + +-/* To estimate if BBR_STARTUP mode (i.e. high_gain) has filled pipe... */ ++/* To estimate if BBR_STARTUP or BBR_BW_PROBE_UP has filled pipe... */ + /* If bw has increased significantly (1.25x), there may be more bw available: */ + static const u32 bbr_full_bw_thresh = BBR_UNIT * 5 / 4; + /* But after 3 rounds w/o significant bw growth, estimate pipe is full: */ + static const u32 bbr_full_bw_cnt = 3; + +-/* "long-term" ("LT") bandwidth estimator parameters... */ +-/* The minimum number of rounds in an LT bw sampling interval: */ +-static const u32 bbr_lt_intvl_min_rtts = 4; +-/* If lost/delivered ratio > 20%, interval is "lossy" and we may be policed: */ +-static const u32 bbr_lt_loss_thresh = 50; +-/* If 2 intervals have a bw ratio <= 1/8, their bw is "consistent": */ +-static const u32 bbr_lt_bw_ratio = BBR_UNIT / 8; +-/* If 2 intervals have a bw diff <= 4 Kbit/sec their bw is "consistent": */ +-static const u32 bbr_lt_bw_diff = 4000 / 8; +-/* If we estimate we're policed, use lt_bw for this many round trips: */ +-static const u32 bbr_lt_bw_max_rtts = 48; +- + /* Gain factor for adding extra_acked to target cwnd: */ + static const int bbr_extra_acked_gain = BBR_UNIT; + /* Window length of extra_acked window. */ +@@ -201,8 +256,121 @@ static const u32 bbr_ack_epoch_acked_reset_thresh = 1U << 20; + /* Time period for clamping cwnd increment due to ack aggregation */ + static const u32 bbr_extra_acked_max_us = 100 * 1000; + ++/* Flags to control BBR ECN-related behavior... */ ++ ++/* Ensure ACKs only ACK packets with consistent ECN CE status? */ ++static const bool bbr_precise_ece_ack = true; ++ ++/* Max RTT (in usec) at which to use sender-side ECN logic. ++ * Disabled when 0 (ECN allowed at any RTT). ++ */ ++static const u32 bbr_ecn_max_rtt_us = 5000; ++ ++/* On losses, scale down inflight and pacing rate by beta scaled by BBR_SCALE. ++ * No loss response when 0. ++ */ ++static const u32 bbr_beta = BBR_UNIT * 30 / 100; ++ ++/* Gain factor for ECN mark ratio samples, scaled by BBR_SCALE (1/16 = 6.25%) */ ++static const u32 bbr_ecn_alpha_gain = BBR_UNIT * 1 / 16; ++ ++/* The initial value for ecn_alpha; 1.0 allows a flow to respond quickly ++ * to congestion if the bottleneck is congested when the flow starts up. ++ */ ++static const u32 bbr_ecn_alpha_init = BBR_UNIT; ++ ++/* On ECN, cut inflight_lo to (1 - ecn_factor * ecn_alpha) scaled by BBR_SCALE. ++ * No ECN based bounding when 0. ++ */ ++static const u32 bbr_ecn_factor = BBR_UNIT * 1 / 3; /* 1/3 = 33% */ ++ ++/* Estimate bw probing has gone too far if CE ratio exceeds this threshold. ++ * Scaled by BBR_SCALE. Disabled when 0. ++ */ ++static const u32 bbr_ecn_thresh = BBR_UNIT * 1 / 2; /* 1/2 = 50% */ ++ ++/* If non-zero, if in a cycle with no losses but some ECN marks, after ECN ++ * clears then make the first round's increment to inflight_hi the following ++ * fraction of inflight_hi. ++ */ ++static const u32 bbr_ecn_reprobe_gain = BBR_UNIT * 1 / 2; ++ ++/* Estimate bw probing has gone too far if loss rate exceeds this level. */ ++static const u32 bbr_loss_thresh = BBR_UNIT * 2 / 100; /* 2% loss */ ++ ++/* Slow down for a packet loss recovered by TLP? */ ++static const bool bbr_loss_probe_recovery = true; ++ ++/* Exit STARTUP if number of loss marking events in a Recovery round is >= N, ++ * and loss rate is higher than bbr_loss_thresh. ++ * Disabled if 0. ++ */ ++static const u32 bbr_full_loss_cnt = 6; ++ ++/* Exit STARTUP if number of round trips with ECN mark rate above ecn_thresh ++ * meets this count. ++ */ ++static const u32 bbr_full_ecn_cnt = 2; ++ ++/* Fraction of unutilized headroom to try to leave in path upon high loss. */ ++static const u32 bbr_inflight_headroom = BBR_UNIT * 15 / 100; ++ ++/* How much do we increase cwnd_gain when probing for bandwidth in ++ * BBR_BW_PROBE_UP? This specifies the increment in units of ++ * BBR_UNIT/4. The default is 1, meaning 0.25. ++ * The min value is 0 (meaning 0.0); max is 3 (meaning 0.75). ++ */ ++static const u32 bbr_bw_probe_cwnd_gain = 1; ++ ++/* Max number of packet-timed rounds to wait before probing for bandwidth. If ++ * we want to tolerate 1% random loss per round, and not have this cut our ++ * inflight too much, we must probe for bw periodically on roughly this scale. ++ * If low, limits Reno/CUBIC coexistence; if high, limits loss tolerance. ++ * We aim to be fair with Reno/CUBIC up to a BDP of at least: ++ * BDP = 25Mbps * .030sec /(1514bytes) = 61.9 packets ++ */ ++static const u32 bbr_bw_probe_max_rounds = 63; ++ ++/* Max amount of randomness to inject in round counting for Reno-coexistence. ++ */ ++static const u32 bbr_bw_probe_rand_rounds = 2; ++ ++/* Use BBR-native probe time scale starting at this many usec. ++ * We aim to be fair with Reno/CUBIC up to an inter-loss time epoch of at least: ++ * BDP*RTT = 25Mbps * .030sec /(1514bytes) * 0.030sec = 1.9 secs ++ */ ++static const u32 bbr_bw_probe_base_us = 2 * USEC_PER_SEC; /* 2 secs */ ++ ++/* Use BBR-native probes spread over this many usec: */ ++static const u32 bbr_bw_probe_rand_us = 1 * USEC_PER_SEC; /* 1 secs */ ++ ++/* Use fast path if app-limited, no loss/ECN, and target cwnd was reached? */ ++static const bool bbr_fast_path = true; ++ ++/* Use fast ack mode? */ ++static const bool bbr_fast_ack_mode = true; ++ ++static u32 bbr_max_bw(const struct sock *sk); ++static u32 bbr_bw(const struct sock *sk); ++static void bbr_exit_probe_rtt(struct sock *sk); ++static void bbr_reset_congestion_signals(struct sock *sk); ++static void bbr_run_loss_probe_recovery(struct sock *sk); ++ + static void bbr_check_probe_rtt_done(struct sock *sk); + ++/* This connection can use ECN if both endpoints have signaled ECN support in ++ * the handshake and the per-route settings indicated this is a ++ * shallow-threshold ECN environment, meaning both: ++ * (a) ECN CE marks indicate low-latency/shallow-threshold congestion, and ++ * (b) TCP endpoints provide precise ACKs that only ACK data segments ++ * with consistent ECN CE status ++ */ ++static bool bbr_can_use_ecn(const struct sock *sk) ++{ ++ return (tcp_sk(sk)->ecn_flags & TCP_ECN_OK) && ++ (tcp_sk(sk)->ecn_flags & TCP_ECN_LOW); ++} ++ + /* Do we estimate that STARTUP filled the pipe? */ + static bool bbr_full_bw_reached(const struct sock *sk) + { +@@ -214,17 +382,17 @@ static bool bbr_full_bw_reached(const struct sock *sk) + /* Return the windowed max recent bandwidth sample, in pkts/uS << BW_SCALE. */ + static u32 bbr_max_bw(const struct sock *sk) + { +- struct bbr *bbr = inet_csk_ca(sk); ++ const struct bbr *bbr = inet_csk_ca(sk); + +- return minmax_get(&bbr->bw); ++ return max(bbr->bw_hi[0], bbr->bw_hi[1]); + } + + /* Return the estimated bandwidth of the path, in pkts/uS << BW_SCALE. */ + static u32 bbr_bw(const struct sock *sk) + { +- struct bbr *bbr = inet_csk_ca(sk); ++ const struct bbr *bbr = inet_csk_ca(sk); + +- return bbr->lt_use_bw ? bbr->lt_bw : bbr_max_bw(sk); ++ return min(bbr_max_bw(sk), bbr->bw_lo); + } + + /* Return maximum extra acked in past k-2k round trips, +@@ -241,15 +409,23 @@ static u16 bbr_extra_acked(const struct sock *sk) + * The order here is chosen carefully to avoid overflow of u64. This should + * work for input rates of up to 2.9Tbit/sec and gain of 2.89x. + */ +-static u64 bbr_rate_bytes_per_sec(struct sock *sk, u64 rate, int gain) ++static u64 bbr_rate_bytes_per_sec(struct sock *sk, u64 rate, int gain, ++ int margin) + { + unsigned int mss = tcp_sk(sk)->mss_cache; + + rate *= mss; + rate *= gain; + rate >>= BBR_SCALE; +- rate *= USEC_PER_SEC / 100 * (100 - bbr_pacing_margin_percent); +- return rate >> BW_SCALE; ++ rate *= USEC_PER_SEC / 100 * (100 - margin); ++ rate >>= BW_SCALE; ++ rate = max(rate, 1ULL); ++ return rate; ++} ++ ++static u64 bbr_bw_bytes_per_sec(struct sock *sk, u64 rate) ++{ ++ return bbr_rate_bytes_per_sec(sk, rate, BBR_UNIT, 0); + } + + /* Convert a BBR bw and gain factor to a pacing rate in bytes per second. */ +@@ -257,12 +433,13 @@ static unsigned long bbr_bw_to_pacing_rate(struct sock *sk, u32 bw, int gain) + { + u64 rate = bw; + +- rate = bbr_rate_bytes_per_sec(sk, rate, gain); ++ rate = bbr_rate_bytes_per_sec(sk, rate, gain, ++ bbr_pacing_margin_percent); + rate = min_t(u64, rate, READ_ONCE(sk->sk_max_pacing_rate)); + return rate; + } + +-/* Initialize pacing rate to: high_gain * init_cwnd / RTT. */ ++/* Initialize pacing rate to: startup_pacing_gain * init_cwnd / RTT. */ + static void bbr_init_pacing_rate_from_rtt(struct sock *sk) + { + struct tcp_sock *tp = tcp_sk(sk); +@@ -279,7 +456,7 @@ static void bbr_init_pacing_rate_from_rtt(struct sock *sk) + bw = (u64)tcp_snd_cwnd(tp) * BW_UNIT; + do_div(bw, rtt_us); + WRITE_ONCE(sk->sk_pacing_rate, +- bbr_bw_to_pacing_rate(sk, bw, bbr_high_gain)); ++ bbr_bw_to_pacing_rate(sk, bw, bbr_param(sk, startup_pacing_gain))); + } + + /* Pace using current bw estimate and a gain factor. */ +@@ -295,26 +472,48 @@ static void bbr_set_pacing_rate(struct sock *sk, u32 bw, int gain) + WRITE_ONCE(sk->sk_pacing_rate, rate); + } + +-/* override sysctl_tcp_min_tso_segs */ +-__bpf_kfunc static u32 bbr_min_tso_segs(struct sock *sk) ++/* Return the number of segments BBR would like in a TSO/GSO skb, given a ++ * particular max gso size as a constraint. TODO: make this simpler and more ++ * consistent by switching bbr to just call tcp_tso_autosize(). ++ */ ++static u32 bbr_tso_segs_generic(struct sock *sk, unsigned int mss_now, ++ u32 gso_max_size) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 segs, r; ++ u64 bytes; ++ ++ /* Budget a TSO/GSO burst size allowance based on bw (pacing_rate). */ ++ bytes = READ_ONCE(sk->sk_pacing_rate) >> READ_ONCE(sk->sk_pacing_shift); ++ ++ /* Budget a TSO/GSO burst size allowance based on min_rtt. For every ++ * K = 2^tso_rtt_shift microseconds of min_rtt, halve the burst. ++ * The min_rtt-based burst allowance is: 64 KBytes / 2^(min_rtt/K) ++ */ ++ if (bbr_param(sk, tso_rtt_shift)) { ++ r = bbr->min_rtt_us >> bbr_param(sk, tso_rtt_shift); ++ if (r < BITS_PER_TYPE(u32)) /* prevent undefined behavior */ ++ bytes += GSO_LEGACY_MAX_SIZE >> r; ++ } ++ ++ bytes = min_t(u32, bytes, gso_max_size - 1 - MAX_TCP_HEADER); ++ segs = max_t(u32, bytes / mss_now, ++ sock_net(sk)->ipv4.sysctl_tcp_min_tso_segs); ++ return segs; ++} ++ ++/* Custom tcp_tso_autosize() for BBR, used at transmit time to cap skb size. */ ++__bpf_kfunc static u32 bbr_tso_segs(struct sock *sk, unsigned int mss_now) + { +- return READ_ONCE(sk->sk_pacing_rate) < (bbr_min_tso_rate >> 3) ? 1 : 2; ++ return bbr_tso_segs_generic(sk, mss_now, sk->sk_gso_max_size); + } + ++/* Like bbr_tso_segs(), using mss_cache, ignoring driver's sk_gso_max_size. */ + static u32 bbr_tso_segs_goal(struct sock *sk) + { + struct tcp_sock *tp = tcp_sk(sk); +- u32 segs, bytes; +- +- /* Sort of tcp_tso_autosize() but ignoring +- * driver provided sk_gso_max_size. +- */ +- bytes = min_t(unsigned long, +- READ_ONCE(sk->sk_pacing_rate) >> READ_ONCE(sk->sk_pacing_shift), +- GSO_LEGACY_MAX_SIZE - 1 - MAX_TCP_HEADER); +- segs = max_t(u32, bytes / tp->mss_cache, bbr_min_tso_segs(sk)); + +- return min(segs, 0x7FU); ++ return bbr_tso_segs_generic(sk, tp->mss_cache, GSO_LEGACY_MAX_SIZE); + } + + /* Save "last known good" cwnd so we can restore it after losses or PROBE_RTT */ +@@ -334,7 +533,9 @@ __bpf_kfunc static void bbr_cwnd_event(struct sock *sk, enum tcp_ca_event event) + struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); + +- if (event == CA_EVENT_TX_START && tp->app_limited) { ++ if (event == CA_EVENT_TX_START) { ++ if (!tp->app_limited) ++ return; + bbr->idle_restart = 1; + bbr->ack_epoch_mstamp = tp->tcp_mstamp; + bbr->ack_epoch_acked = 0; +@@ -345,6 +546,16 @@ __bpf_kfunc static void bbr_cwnd_event(struct sock *sk, enum tcp_ca_event event) + bbr_set_pacing_rate(sk, bbr_bw(sk), BBR_UNIT); + else if (bbr->mode == BBR_PROBE_RTT) + bbr_check_probe_rtt_done(sk); ++ } else if ((event == CA_EVENT_ECN_IS_CE || ++ event == CA_EVENT_ECN_NO_CE) && ++ bbr_can_use_ecn(sk) && ++ bbr_param(sk, precise_ece_ack)) { ++ u32 state = bbr->ce_state; ++ dctcp_ece_ack_update(sk, event, &bbr->prior_rcv_nxt, &state); ++ bbr->ce_state = state; ++ } else if (event == CA_EVENT_TLP_RECOVERY && ++ bbr_param(sk, loss_probe_recovery)) { ++ bbr_run_loss_probe_recovery(sk); + } + } + +@@ -367,10 +578,10 @@ static u32 bbr_bdp(struct sock *sk, u32 bw, int gain) + * default. This should only happen when the connection is not using TCP + * timestamps and has retransmitted all of the SYN/SYNACK/data packets + * ACKed so far. In this case, an RTO can cut cwnd to 1, in which +- * case we need to slow-start up toward something safe: TCP_INIT_CWND. ++ * case we need to slow-start up toward something safe: initial cwnd. + */ + if (unlikely(bbr->min_rtt_us == ~0U)) /* no valid RTT samples yet? */ +- return TCP_INIT_CWND; /* be safe: cap at default initial cwnd*/ ++ return bbr->init_cwnd; /* be safe: cap at initial cwnd */ + + w = (u64)bw * bbr->min_rtt_us; + +@@ -387,23 +598,23 @@ static u32 bbr_bdp(struct sock *sk, u32 bw, int gain) + * - one skb in sending host Qdisc, + * - one skb in sending host TSO/GSO engine + * - one skb being received by receiver host LRO/GRO/delayed-ACK engine +- * Don't worry, at low rates (bbr_min_tso_rate) this won't bloat cwnd because +- * in such cases tso_segs_goal is 1. The minimum cwnd is 4 packets, ++ * Don't worry, at low rates this won't bloat cwnd because ++ * in such cases tso_segs_goal is small. The minimum cwnd is 4 packets, + * which allows 2 outstanding 2-packet sequences, to try to keep pipe + * full even with ACK-every-other-packet delayed ACKs. + */ + static u32 bbr_quantization_budget(struct sock *sk, u32 cwnd) + { + struct bbr *bbr = inet_csk_ca(sk); ++ u32 tso_segs_goal; + +- /* Allow enough full-sized skbs in flight to utilize end systems. */ +- cwnd += 3 * bbr_tso_segs_goal(sk); +- +- /* Reduce delayed ACKs by rounding up cwnd to the next even number. */ +- cwnd = (cwnd + 1) & ~1U; ++ tso_segs_goal = 3 * bbr_tso_segs_goal(sk); + ++ /* Allow enough full-sized skbs in flight to utilize end systems. */ ++ cwnd = max_t(u32, cwnd, tso_segs_goal); ++ cwnd = max_t(u32, cwnd, bbr_param(sk, cwnd_min_target)); + /* Ensure gain cycling gets inflight above BDP even for small BDPs. */ +- if (bbr->mode == BBR_PROBE_BW && bbr->cycle_idx == 0) ++ if (bbr->mode == BBR_PROBE_BW && bbr->cycle_idx == BBR_BW_PROBE_UP) + cwnd += 2; + + return cwnd; +@@ -458,10 +669,10 @@ static u32 bbr_ack_aggregation_cwnd(struct sock *sk) + { + u32 max_aggr_cwnd, aggr_cwnd = 0; + +- if (bbr_extra_acked_gain && bbr_full_bw_reached(sk)) { ++ if (bbr_param(sk, extra_acked_gain)) { + max_aggr_cwnd = ((u64)bbr_bw(sk) * bbr_extra_acked_max_us) + / BW_UNIT; +- aggr_cwnd = (bbr_extra_acked_gain * bbr_extra_acked(sk)) ++ aggr_cwnd = (bbr_param(sk, extra_acked_gain) * bbr_extra_acked(sk)) + >> BBR_SCALE; + aggr_cwnd = min(aggr_cwnd, max_aggr_cwnd); + } +@@ -469,66 +680,27 @@ static u32 bbr_ack_aggregation_cwnd(struct sock *sk) + return aggr_cwnd; + } + +-/* An optimization in BBR to reduce losses: On the first round of recovery, we +- * follow the packet conservation principle: send P packets per P packets acked. +- * After that, we slow-start and send at most 2*P packets per P packets acked. +- * After recovery finishes, or upon undo, we restore the cwnd we had when +- * recovery started (capped by the target cwnd based on estimated BDP). +- * +- * TODO(ycheng/ncardwell): implement a rate-based approach. +- */ +-static bool bbr_set_cwnd_to_recover_or_restore( +- struct sock *sk, const struct rate_sample *rs, u32 acked, u32 *new_cwnd) ++/* Returns the cwnd for PROBE_RTT mode. */ ++static u32 bbr_probe_rtt_cwnd(struct sock *sk) + { +- struct tcp_sock *tp = tcp_sk(sk); +- struct bbr *bbr = inet_csk_ca(sk); +- u8 prev_state = bbr->prev_ca_state, state = inet_csk(sk)->icsk_ca_state; +- u32 cwnd = tcp_snd_cwnd(tp); +- +- /* An ACK for P pkts should release at most 2*P packets. We do this +- * in two steps. First, here we deduct the number of lost packets. +- * Then, in bbr_set_cwnd() we slow start up toward the target cwnd. +- */ +- if (rs->losses > 0) +- cwnd = max_t(s32, cwnd - rs->losses, 1); +- +- if (state == TCP_CA_Recovery && prev_state != TCP_CA_Recovery) { +- /* Starting 1st round of Recovery, so do packet conservation. */ +- bbr->packet_conservation = 1; +- bbr->next_rtt_delivered = tp->delivered; /* start round now */ +- /* Cut unused cwnd from app behavior, TSQ, or TSO deferral: */ +- cwnd = tcp_packets_in_flight(tp) + acked; +- } else if (prev_state >= TCP_CA_Recovery && state < TCP_CA_Recovery) { +- /* Exiting loss recovery; restore cwnd saved before recovery. */ +- cwnd = max(cwnd, bbr->prior_cwnd); +- bbr->packet_conservation = 0; +- } +- bbr->prev_ca_state = state; +- +- if (bbr->packet_conservation) { +- *new_cwnd = max(cwnd, tcp_packets_in_flight(tp) + acked); +- return true; /* yes, using packet conservation */ +- } +- *new_cwnd = cwnd; +- return false; ++ return max_t(u32, bbr_param(sk, cwnd_min_target), ++ bbr_bdp(sk, bbr_bw(sk), bbr_param(sk, probe_rtt_cwnd_gain))); + } + + /* Slow-start up toward target cwnd (if bw estimate is growing, or packet loss + * has drawn us down below target), or snap down to target if we're above it. + */ + static void bbr_set_cwnd(struct sock *sk, const struct rate_sample *rs, +- u32 acked, u32 bw, int gain) ++ u32 acked, u32 bw, int gain, u32 cwnd, ++ struct bbr_context *ctx) + { + struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); +- u32 cwnd = tcp_snd_cwnd(tp), target_cwnd = 0; ++ u32 target_cwnd = 0; + + if (!acked) + goto done; /* no packet fully ACKed; just apply caps */ + +- if (bbr_set_cwnd_to_recover_or_restore(sk, rs, acked, &cwnd)) +- goto done; +- + target_cwnd = bbr_bdp(sk, bw, gain); + + /* Increment the cwnd to account for excess ACKed data that seems +@@ -537,74 +709,26 @@ static void bbr_set_cwnd(struct sock *sk, const struct rate_sample *rs, + target_cwnd += bbr_ack_aggregation_cwnd(sk); + target_cwnd = bbr_quantization_budget(sk, target_cwnd); + +- /* If we're below target cwnd, slow start cwnd toward target cwnd. */ +- if (bbr_full_bw_reached(sk)) /* only cut cwnd if we filled the pipe */ +- cwnd = min(cwnd + acked, target_cwnd); +- else if (cwnd < target_cwnd || tp->delivered < TCP_INIT_CWND) +- cwnd = cwnd + acked; +- cwnd = max(cwnd, bbr_cwnd_min_target); ++ /* Update cwnd and enable fast path if cwnd reaches target_cwnd. */ ++ bbr->try_fast_path = 0; ++ if (bbr_full_bw_reached(sk)) { /* only cut cwnd if we filled the pipe */ ++ cwnd += acked; ++ if (cwnd >= target_cwnd) { ++ cwnd = target_cwnd; ++ bbr->try_fast_path = 1; ++ } ++ } else if (cwnd < target_cwnd || cwnd < 2 * bbr->init_cwnd) { ++ cwnd += acked; ++ } else { ++ bbr->try_fast_path = 1; ++ } + ++ cwnd = max_t(u32, cwnd, bbr_param(sk, cwnd_min_target)); + done: +- tcp_snd_cwnd_set(tp, min(cwnd, tp->snd_cwnd_clamp)); /* apply global cap */ ++ tcp_snd_cwnd_set(tp, min(cwnd, tp->snd_cwnd_clamp)); /* global cap */ + if (bbr->mode == BBR_PROBE_RTT) /* drain queue, refresh min_rtt */ +- tcp_snd_cwnd_set(tp, min(tcp_snd_cwnd(tp), bbr_cwnd_min_target)); +-} +- +-/* End cycle phase if it's time and/or we hit the phase's in-flight target. */ +-static bool bbr_is_next_cycle_phase(struct sock *sk, +- const struct rate_sample *rs) +-{ +- struct tcp_sock *tp = tcp_sk(sk); +- struct bbr *bbr = inet_csk_ca(sk); +- bool is_full_length = +- tcp_stamp_us_delta(tp->delivered_mstamp, bbr->cycle_mstamp) > +- bbr->min_rtt_us; +- u32 inflight, bw; +- +- /* The pacing_gain of 1.0 paces at the estimated bw to try to fully +- * use the pipe without increasing the queue. +- */ +- if (bbr->pacing_gain == BBR_UNIT) +- return is_full_length; /* just use wall clock time */ +- +- inflight = bbr_packets_in_net_at_edt(sk, rs->prior_in_flight); +- bw = bbr_max_bw(sk); +- +- /* A pacing_gain > 1.0 probes for bw by trying to raise inflight to at +- * least pacing_gain*BDP; this may take more than min_rtt if min_rtt is +- * small (e.g. on a LAN). We do not persist if packets are lost, since +- * a path with small buffers may not hold that much. +- */ +- if (bbr->pacing_gain > BBR_UNIT) +- return is_full_length && +- (rs->losses || /* perhaps pacing_gain*BDP won't fit */ +- inflight >= bbr_inflight(sk, bw, bbr->pacing_gain)); +- +- /* A pacing_gain < 1.0 tries to drain extra queue we added if bw +- * probing didn't find more bw. If inflight falls to match BDP then we +- * estimate queue is drained; persisting would underutilize the pipe. +- */ +- return is_full_length || +- inflight <= bbr_inflight(sk, bw, BBR_UNIT); +-} +- +-static void bbr_advance_cycle_phase(struct sock *sk) +-{ +- struct tcp_sock *tp = tcp_sk(sk); +- struct bbr *bbr = inet_csk_ca(sk); +- +- bbr->cycle_idx = (bbr->cycle_idx + 1) & (CYCLE_LEN - 1); +- bbr->cycle_mstamp = tp->delivered_mstamp; +-} +- +-/* Gain cycling: cycle pacing gain to converge to fair share of available bw. */ +-static void bbr_update_cycle_phase(struct sock *sk, +- const struct rate_sample *rs) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- +- if (bbr->mode == BBR_PROBE_BW && bbr_is_next_cycle_phase(sk, rs)) +- bbr_advance_cycle_phase(sk); ++ tcp_snd_cwnd_set(tp, min_t(u32, tcp_snd_cwnd(tp), ++ bbr_probe_rtt_cwnd(sk))); + } + + static void bbr_reset_startup_mode(struct sock *sk) +@@ -614,191 +738,49 @@ static void bbr_reset_startup_mode(struct sock *sk) + bbr->mode = BBR_STARTUP; + } + +-static void bbr_reset_probe_bw_mode(struct sock *sk) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- +- bbr->mode = BBR_PROBE_BW; +- bbr->cycle_idx = CYCLE_LEN - 1 - get_random_u32_below(bbr_cycle_rand); +- bbr_advance_cycle_phase(sk); /* flip to next phase of gain cycle */ +-} +- +-static void bbr_reset_mode(struct sock *sk) +-{ +- if (!bbr_full_bw_reached(sk)) +- bbr_reset_startup_mode(sk); +- else +- bbr_reset_probe_bw_mode(sk); +-} +- +-/* Start a new long-term sampling interval. */ +-static void bbr_reset_lt_bw_sampling_interval(struct sock *sk) +-{ +- struct tcp_sock *tp = tcp_sk(sk); +- struct bbr *bbr = inet_csk_ca(sk); +- +- bbr->lt_last_stamp = div_u64(tp->delivered_mstamp, USEC_PER_MSEC); +- bbr->lt_last_delivered = tp->delivered; +- bbr->lt_last_lost = tp->lost; +- bbr->lt_rtt_cnt = 0; +-} +- +-/* Completely reset long-term bandwidth sampling. */ +-static void bbr_reset_lt_bw_sampling(struct sock *sk) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- +- bbr->lt_bw = 0; +- bbr->lt_use_bw = 0; +- bbr->lt_is_sampling = false; +- bbr_reset_lt_bw_sampling_interval(sk); +-} +- +-/* Long-term bw sampling interval is done. Estimate whether we're policed. */ +-static void bbr_lt_bw_interval_done(struct sock *sk, u32 bw) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- u32 diff; +- +- if (bbr->lt_bw) { /* do we have bw from a previous interval? */ +- /* Is new bw close to the lt_bw from the previous interval? */ +- diff = abs(bw - bbr->lt_bw); +- if ((diff * BBR_UNIT <= bbr_lt_bw_ratio * bbr->lt_bw) || +- (bbr_rate_bytes_per_sec(sk, diff, BBR_UNIT) <= +- bbr_lt_bw_diff)) { +- /* All criteria are met; estimate we're policed. */ +- bbr->lt_bw = (bw + bbr->lt_bw) >> 1; /* avg 2 intvls */ +- bbr->lt_use_bw = 1; +- bbr->pacing_gain = BBR_UNIT; /* try to avoid drops */ +- bbr->lt_rtt_cnt = 0; +- return; +- } +- } +- bbr->lt_bw = bw; +- bbr_reset_lt_bw_sampling_interval(sk); +-} +- +-/* Token-bucket traffic policers are common (see "An Internet-Wide Analysis of +- * Traffic Policing", SIGCOMM 2016). BBR detects token-bucket policers and +- * explicitly models their policed rate, to reduce unnecessary losses. We +- * estimate that we're policed if we see 2 consecutive sampling intervals with +- * consistent throughput and high packet loss. If we think we're being policed, +- * set lt_bw to the "long-term" average delivery rate from those 2 intervals. ++/* See if we have reached next round trip. Upon start of the new round, ++ * returns packets delivered since previous round start plus this ACK. + */ +-static void bbr_lt_bw_sampling(struct sock *sk, const struct rate_sample *rs) +-{ +- struct tcp_sock *tp = tcp_sk(sk); +- struct bbr *bbr = inet_csk_ca(sk); +- u32 lost, delivered; +- u64 bw; +- u32 t; +- +- if (bbr->lt_use_bw) { /* already using long-term rate, lt_bw? */ +- if (bbr->mode == BBR_PROBE_BW && bbr->round_start && +- ++bbr->lt_rtt_cnt >= bbr_lt_bw_max_rtts) { +- bbr_reset_lt_bw_sampling(sk); /* stop using lt_bw */ +- bbr_reset_probe_bw_mode(sk); /* restart gain cycling */ +- } +- return; +- } +- +- /* Wait for the first loss before sampling, to let the policer exhaust +- * its tokens and estimate the steady-state rate allowed by the policer. +- * Starting samples earlier includes bursts that over-estimate the bw. +- */ +- if (!bbr->lt_is_sampling) { +- if (!rs->losses) +- return; +- bbr_reset_lt_bw_sampling_interval(sk); +- bbr->lt_is_sampling = true; +- } +- +- /* To avoid underestimates, reset sampling if we run out of data. */ +- if (rs->is_app_limited) { +- bbr_reset_lt_bw_sampling(sk); +- return; +- } +- +- if (bbr->round_start) +- bbr->lt_rtt_cnt++; /* count round trips in this interval */ +- if (bbr->lt_rtt_cnt < bbr_lt_intvl_min_rtts) +- return; /* sampling interval needs to be longer */ +- if (bbr->lt_rtt_cnt > 4 * bbr_lt_intvl_min_rtts) { +- bbr_reset_lt_bw_sampling(sk); /* interval is too long */ +- return; +- } +- +- /* End sampling interval when a packet is lost, so we estimate the +- * policer tokens were exhausted. Stopping the sampling before the +- * tokens are exhausted under-estimates the policed rate. +- */ +- if (!rs->losses) +- return; +- +- /* Calculate packets lost and delivered in sampling interval. */ +- lost = tp->lost - bbr->lt_last_lost; +- delivered = tp->delivered - bbr->lt_last_delivered; +- /* Is loss rate (lost/delivered) >= lt_loss_thresh? If not, wait. */ +- if (!delivered || (lost << BBR_SCALE) < bbr_lt_loss_thresh * delivered) +- return; +- +- /* Find average delivery rate in this sampling interval. */ +- t = div_u64(tp->delivered_mstamp, USEC_PER_MSEC) - bbr->lt_last_stamp; +- if ((s32)t < 1) +- return; /* interval is less than one ms, so wait */ +- /* Check if can multiply without overflow */ +- if (t >= ~0U / USEC_PER_MSEC) { +- bbr_reset_lt_bw_sampling(sk); /* interval too long; reset */ +- return; +- } +- t *= USEC_PER_MSEC; +- bw = (u64)delivered * BW_UNIT; +- do_div(bw, t); +- bbr_lt_bw_interval_done(sk, bw); +-} +- +-/* Estimate the bandwidth based on how fast packets are delivered */ +-static void bbr_update_bw(struct sock *sk, const struct rate_sample *rs) ++static u32 bbr_update_round_start(struct sock *sk, ++ const struct rate_sample *rs, struct bbr_context *ctx) + { + struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); +- u64 bw; ++ u32 round_delivered = 0; + + bbr->round_start = 0; +- if (rs->delivered < 0 || rs->interval_us <= 0) +- return; /* Not a valid observation */ + + /* See if we've reached the next RTT */ +- if (!before(rs->prior_delivered, bbr->next_rtt_delivered)) { ++ if (rs->interval_us > 0 && ++ !before(rs->prior_delivered, bbr->next_rtt_delivered)) { ++ round_delivered = tp->delivered - bbr->next_rtt_delivered; + bbr->next_rtt_delivered = tp->delivered; +- bbr->rtt_cnt++; + bbr->round_start = 1; +- bbr->packet_conservation = 0; + } ++ return round_delivered; ++} + +- bbr_lt_bw_sampling(sk, rs); ++/* Calculate the bandwidth based on how fast packets are delivered */ ++static void bbr_calculate_bw_sample(struct sock *sk, ++ const struct rate_sample *rs, struct bbr_context *ctx) ++{ ++ u64 bw = 0; + + /* Divide delivered by the interval to find a (lower bound) bottleneck + * bandwidth sample. Delivered is in packets and interval_us in uS and + * ratio will be <<1 for most connections. So delivered is first scaled. ++ * Round up to allow growth at low rates, even with integer division. + */ +- bw = div64_long((u64)rs->delivered * BW_UNIT, rs->interval_us); +- +- /* If this sample is application-limited, it is likely to have a very +- * low delivered count that represents application behavior rather than +- * the available network rate. Such a sample could drag down estimated +- * bw, causing needless slow-down. Thus, to continue to send at the +- * last measured network rate, we filter out app-limited samples unless +- * they describe the path bw at least as well as our bw model. +- * +- * So the goal during app-limited phase is to proceed with the best +- * network rate no matter how long. We automatically leave this +- * phase when app writes faster than the network can deliver :) +- */ +- if (!rs->is_app_limited || bw >= bbr_max_bw(sk)) { +- /* Incorporate new sample into our max bw filter. */ +- minmax_running_max(&bbr->bw, bbr_bw_rtts, bbr->rtt_cnt, bw); ++ if (rs->interval_us > 0) { ++ if (WARN_ONCE(rs->delivered < 0, ++ "negative delivered: %d interval_us: %ld\n", ++ rs->delivered, rs->interval_us)) ++ return; ++ ++ bw = DIV_ROUND_UP_ULL((u64)rs->delivered * BW_UNIT, rs->interval_us); + } ++ ++ ctx->sample_bw = bw; + } + + /* Estimates the windowed max degree of ack aggregation. +@@ -812,7 +794,7 @@ static void bbr_update_bw(struct sock *sk, const struct rate_sample *rs) + * + * Max extra_acked is clamped by cwnd and bw * bbr_extra_acked_max_us (100 ms). + * Max filter is an approximate sliding window of 5-10 (packet timed) round +- * trips. ++ * trips for non-startup phase, and 1-2 round trips for startup. + */ + static void bbr_update_ack_aggregation(struct sock *sk, + const struct rate_sample *rs) +@@ -820,15 +802,19 @@ static void bbr_update_ack_aggregation(struct sock *sk, + u32 epoch_us, expected_acked, extra_acked; + struct bbr *bbr = inet_csk_ca(sk); + struct tcp_sock *tp = tcp_sk(sk); ++ u32 extra_acked_win_rtts_thresh = bbr_param(sk, extra_acked_win_rtts); + +- if (!bbr_extra_acked_gain || rs->acked_sacked <= 0 || ++ if (!bbr_param(sk, extra_acked_gain) || rs->acked_sacked <= 0 || + rs->delivered < 0 || rs->interval_us <= 0) + return; + + if (bbr->round_start) { + bbr->extra_acked_win_rtts = min(0x1F, + bbr->extra_acked_win_rtts + 1); +- if (bbr->extra_acked_win_rtts >= bbr_extra_acked_win_rtts) { ++ if (!bbr_full_bw_reached(sk)) ++ extra_acked_win_rtts_thresh = 1; ++ if (bbr->extra_acked_win_rtts >= ++ extra_acked_win_rtts_thresh) { + bbr->extra_acked_win_rtts = 0; + bbr->extra_acked_win_idx = bbr->extra_acked_win_idx ? + 0 : 1; +@@ -862,49 +848,6 @@ static void bbr_update_ack_aggregation(struct sock *sk, + bbr->extra_acked[bbr->extra_acked_win_idx] = extra_acked; + } + +-/* Estimate when the pipe is full, using the change in delivery rate: BBR +- * estimates that STARTUP filled the pipe if the estimated bw hasn't changed by +- * at least bbr_full_bw_thresh (25%) after bbr_full_bw_cnt (3) non-app-limited +- * rounds. Why 3 rounds: 1: rwin autotuning grows the rwin, 2: we fill the +- * higher rwin, 3: we get higher delivery rate samples. Or transient +- * cross-traffic or radio noise can go away. CUBIC Hystart shares a similar +- * design goal, but uses delay and inter-ACK spacing instead of bandwidth. +- */ +-static void bbr_check_full_bw_reached(struct sock *sk, +- const struct rate_sample *rs) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- u32 bw_thresh; +- +- if (bbr_full_bw_reached(sk) || !bbr->round_start || rs->is_app_limited) +- return; +- +- bw_thresh = (u64)bbr->full_bw * bbr_full_bw_thresh >> BBR_SCALE; +- if (bbr_max_bw(sk) >= bw_thresh) { +- bbr->full_bw = bbr_max_bw(sk); +- bbr->full_bw_cnt = 0; +- return; +- } +- ++bbr->full_bw_cnt; +- bbr->full_bw_reached = bbr->full_bw_cnt >= bbr_full_bw_cnt; +-} +- +-/* If pipe is probably full, drain the queue and then enter steady-state. */ +-static void bbr_check_drain(struct sock *sk, const struct rate_sample *rs) +-{ +- struct bbr *bbr = inet_csk_ca(sk); +- +- if (bbr->mode == BBR_STARTUP && bbr_full_bw_reached(sk)) { +- bbr->mode = BBR_DRAIN; /* drain queue we created */ +- tcp_sk(sk)->snd_ssthresh = +- bbr_inflight(sk, bbr_max_bw(sk), BBR_UNIT); +- } /* fall through to check if in-flight is already small: */ +- if (bbr->mode == BBR_DRAIN && +- bbr_packets_in_net_at_edt(sk, tcp_packets_in_flight(tcp_sk(sk))) <= +- bbr_inflight(sk, bbr_max_bw(sk), BBR_UNIT)) +- bbr_reset_probe_bw_mode(sk); /* we estimate queue is drained */ +-} +- + static void bbr_check_probe_rtt_done(struct sock *sk) + { + struct tcp_sock *tp = tcp_sk(sk); +@@ -914,9 +857,9 @@ static void bbr_check_probe_rtt_done(struct sock *sk) + after(tcp_jiffies32, bbr->probe_rtt_done_stamp))) + return; + +- bbr->min_rtt_stamp = tcp_jiffies32; /* wait a while until PROBE_RTT */ ++ bbr->probe_rtt_min_stamp = tcp_jiffies32; /* schedule next PROBE_RTT */ + tcp_snd_cwnd_set(tp, max(tcp_snd_cwnd(tp), bbr->prior_cwnd)); +- bbr_reset_mode(sk); ++ bbr_exit_probe_rtt(sk); + } + + /* The goal of PROBE_RTT mode is to have BBR flows cooperatively and +@@ -942,23 +885,35 @@ static void bbr_update_min_rtt(struct sock *sk, const struct rate_sample *rs) + { + struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); +- bool filter_expired; ++ bool probe_rtt_expired, min_rtt_expired; ++ u32 expire; + +- /* Track min RTT seen in the min_rtt_win_sec filter window: */ +- filter_expired = after(tcp_jiffies32, +- bbr->min_rtt_stamp + bbr_min_rtt_win_sec * HZ); ++ /* Track min RTT in probe_rtt_win_ms to time next PROBE_RTT state. */ ++ expire = bbr->probe_rtt_min_stamp + ++ msecs_to_jiffies(bbr_param(sk, probe_rtt_win_ms)); ++ probe_rtt_expired = after(tcp_jiffies32, expire); + if (rs->rtt_us >= 0 && +- (rs->rtt_us < bbr->min_rtt_us || +- (filter_expired && !rs->is_ack_delayed))) { +- bbr->min_rtt_us = rs->rtt_us; +- bbr->min_rtt_stamp = tcp_jiffies32; ++ (rs->rtt_us < bbr->probe_rtt_min_us || ++ (probe_rtt_expired && !rs->is_ack_delayed))) { ++ bbr->probe_rtt_min_us = rs->rtt_us; ++ bbr->probe_rtt_min_stamp = tcp_jiffies32; ++ } ++ /* Track min RTT seen in the min_rtt_win_sec filter window: */ ++ expire = bbr->min_rtt_stamp + bbr_param(sk, min_rtt_win_sec) * HZ; ++ min_rtt_expired = after(tcp_jiffies32, expire); ++ if (bbr->probe_rtt_min_us <= bbr->min_rtt_us || ++ min_rtt_expired) { ++ bbr->min_rtt_us = bbr->probe_rtt_min_us; ++ bbr->min_rtt_stamp = bbr->probe_rtt_min_stamp; + } + +- if (bbr_probe_rtt_mode_ms > 0 && filter_expired && ++ if (bbr_param(sk, probe_rtt_mode_ms) > 0 && probe_rtt_expired && + !bbr->idle_restart && bbr->mode != BBR_PROBE_RTT) { + bbr->mode = BBR_PROBE_RTT; /* dip, drain queue */ + bbr_save_cwnd(sk); /* note cwnd so we can restore it */ + bbr->probe_rtt_done_stamp = 0; ++ bbr->ack_phase = BBR_ACKS_PROBE_STOPPING; ++ bbr->next_rtt_delivered = tp->delivered; + } + + if (bbr->mode == BBR_PROBE_RTT) { +@@ -967,9 +922,9 @@ static void bbr_update_min_rtt(struct sock *sk, const struct rate_sample *rs) + (tp->delivered + tcp_packets_in_flight(tp)) ? : 1; + /* Maintain min packets in flight for max(200 ms, 1 round). */ + if (!bbr->probe_rtt_done_stamp && +- tcp_packets_in_flight(tp) <= bbr_cwnd_min_target) { ++ tcp_packets_in_flight(tp) <= bbr_probe_rtt_cwnd(sk)) { + bbr->probe_rtt_done_stamp = tcp_jiffies32 + +- msecs_to_jiffies(bbr_probe_rtt_mode_ms); ++ msecs_to_jiffies(bbr_param(sk, probe_rtt_mode_ms)); + bbr->probe_rtt_round_done = 0; + bbr->next_rtt_delivered = tp->delivered; + } else if (bbr->probe_rtt_done_stamp) { +@@ -990,18 +945,20 @@ static void bbr_update_gains(struct sock *sk) + + switch (bbr->mode) { + case BBR_STARTUP: +- bbr->pacing_gain = bbr_high_gain; +- bbr->cwnd_gain = bbr_high_gain; ++ bbr->pacing_gain = bbr_param(sk, startup_pacing_gain); ++ bbr->cwnd_gain = bbr_param(sk, startup_cwnd_gain); + break; + case BBR_DRAIN: +- bbr->pacing_gain = bbr_drain_gain; /* slow, to drain */ +- bbr->cwnd_gain = bbr_high_gain; /* keep cwnd */ ++ bbr->pacing_gain = bbr_param(sk, drain_gain); /* slow, to drain */ ++ bbr->cwnd_gain = bbr_param(sk, startup_cwnd_gain); /* keep cwnd */ + break; + case BBR_PROBE_BW: +- bbr->pacing_gain = (bbr->lt_use_bw ? +- BBR_UNIT : +- bbr_pacing_gain[bbr->cycle_idx]); +- bbr->cwnd_gain = bbr_cwnd_gain; ++ bbr->pacing_gain = bbr_pacing_gain[bbr->cycle_idx]; ++ bbr->cwnd_gain = bbr_param(sk, cwnd_gain); ++ if (bbr_param(sk, bw_probe_cwnd_gain) && ++ bbr->cycle_idx == BBR_BW_PROBE_UP) ++ bbr->cwnd_gain += ++ BBR_UNIT * bbr_param(sk, bw_probe_cwnd_gain) / 4; + break; + case BBR_PROBE_RTT: + bbr->pacing_gain = BBR_UNIT; +@@ -1013,144 +970,1387 @@ static void bbr_update_gains(struct sock *sk) + } + } + +-static void bbr_update_model(struct sock *sk, const struct rate_sample *rs) ++__bpf_kfunc static u32 bbr_sndbuf_expand(struct sock *sk) + { +- bbr_update_bw(sk, rs); +- bbr_update_ack_aggregation(sk, rs); +- bbr_update_cycle_phase(sk, rs); +- bbr_check_full_bw_reached(sk, rs); +- bbr_check_drain(sk, rs); +- bbr_update_min_rtt(sk, rs); +- bbr_update_gains(sk); ++ /* Provision 3 * cwnd since BBR may slow-start even during recovery. */ ++ return 3; + } + +-__bpf_kfunc static void bbr_main(struct sock *sk, u32 ack, int flag, const struct rate_sample *rs) ++/* Incorporate a new bw sample into the current window of our max filter. */ ++static void bbr_take_max_bw_sample(struct sock *sk, u32 bw) + { + struct bbr *bbr = inet_csk_ca(sk); +- u32 bw; +- +- bbr_update_model(sk, rs); + +- bw = bbr_bw(sk); +- bbr_set_pacing_rate(sk, bw, bbr->pacing_gain); +- bbr_set_cwnd(sk, rs, rs->acked_sacked, bw, bbr->cwnd_gain); ++ bbr->bw_hi[1] = max(bw, bbr->bw_hi[1]); + } + +-__bpf_kfunc static void bbr_init(struct sock *sk) ++/* Keep max of last 1-2 cycles. Each PROBE_BW cycle, flip filter window. */ ++static void bbr_advance_max_bw_filter(struct sock *sk) + { +- struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); + +- bbr->prior_cwnd = 0; +- tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; +- bbr->rtt_cnt = 0; +- bbr->next_rtt_delivered = tp->delivered; +- bbr->prev_ca_state = TCP_CA_Open; +- bbr->packet_conservation = 0; +- +- bbr->probe_rtt_done_stamp = 0; +- bbr->probe_rtt_round_done = 0; +- bbr->min_rtt_us = tcp_min_rtt(tp); +- bbr->min_rtt_stamp = tcp_jiffies32; +- +- minmax_reset(&bbr->bw, bbr->rtt_cnt, 0); /* init max bw to 0 */ ++ if (!bbr->bw_hi[1]) ++ return; /* no samples in this window; remember old window */ ++ bbr->bw_hi[0] = bbr->bw_hi[1]; ++ bbr->bw_hi[1] = 0; ++} + +- bbr->has_seen_rtt = 0; +- bbr_init_pacing_rate_from_rtt(sk); ++/* Reset the estimator for reaching full bandwidth based on bw plateau. */ ++static void bbr_reset_full_bw(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); + +- bbr->round_start = 0; +- bbr->idle_restart = 0; +- bbr->full_bw_reached = 0; + bbr->full_bw = 0; + bbr->full_bw_cnt = 0; +- bbr->cycle_mstamp = 0; +- bbr->cycle_idx = 0; +- bbr_reset_lt_bw_sampling(sk); +- bbr_reset_startup_mode(sk); ++ bbr->full_bw_now = 0; ++} + +- bbr->ack_epoch_mstamp = tp->tcp_mstamp; +- bbr->ack_epoch_acked = 0; +- bbr->extra_acked_win_rtts = 0; +- bbr->extra_acked_win_idx = 0; +- bbr->extra_acked[0] = 0; +- bbr->extra_acked[1] = 0; ++/* How much do we want in flight? Our BDP, unless congestion cut cwnd. */ ++static u32 bbr_target_inflight(struct sock *sk) ++{ ++ u32 bdp = bbr_inflight(sk, bbr_bw(sk), BBR_UNIT); + +- cmpxchg(&sk->sk_pacing_status, SK_PACING_NONE, SK_PACING_NEEDED); ++ return min(bdp, tcp_sk(sk)->snd_cwnd); + } + +-__bpf_kfunc static u32 bbr_sndbuf_expand(struct sock *sk) ++static bool bbr_is_probing_bandwidth(struct sock *sk) + { +- /* Provision 3 * cwnd since BBR may slow-start even during recovery. */ +- return 3; ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ return (bbr->mode == BBR_STARTUP) || ++ (bbr->mode == BBR_PROBE_BW && ++ (bbr->cycle_idx == BBR_BW_PROBE_REFILL || ++ bbr->cycle_idx == BBR_BW_PROBE_UP)); ++} ++ ++/* Has the given amount of time elapsed since we marked the phase start? */ ++static bool bbr_has_elapsed_in_phase(const struct sock *sk, u32 interval_us) ++{ ++ const struct tcp_sock *tp = tcp_sk(sk); ++ const struct bbr *bbr = inet_csk_ca(sk); ++ ++ return tcp_stamp_us_delta(tp->tcp_mstamp, ++ bbr->cycle_mstamp + interval_us) > 0; ++} ++ ++static void bbr_handle_queue_too_high_in_startup(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 bdp; /* estimated BDP in packets, with quantization budget */ ++ ++ bbr->full_bw_reached = 1; ++ ++ bdp = bbr_inflight(sk, bbr_max_bw(sk), BBR_UNIT); ++ bbr->inflight_hi = max(bdp, bbr->inflight_latest); ++} ++ ++/* Exit STARTUP upon N consecutive rounds with ECN mark rate > ecn_thresh. */ ++static void bbr_check_ecn_too_high_in_startup(struct sock *sk, u32 ce_ratio) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (bbr_full_bw_reached(sk) || !bbr->ecn_eligible || ++ !bbr_param(sk, full_ecn_cnt) || !bbr_param(sk, ecn_thresh)) ++ return; ++ ++ if (ce_ratio >= bbr_param(sk, ecn_thresh)) ++ bbr->startup_ecn_rounds++; ++ else ++ bbr->startup_ecn_rounds = 0; ++ ++ if (bbr->startup_ecn_rounds >= bbr_param(sk, full_ecn_cnt)) { ++ bbr_handle_queue_too_high_in_startup(sk); ++ return; ++ } ++} ++ ++/* Updates ecn_alpha and returns ce_ratio. -1 if not available. */ ++static int bbr_update_ecn_alpha(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct net *net = sock_net(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ s32 delivered, delivered_ce; ++ u64 alpha, ce_ratio; ++ u32 gain; ++ bool want_ecn_alpha; ++ ++ /* See if we should use ECN sender logic for this connection. */ ++ if (!bbr->ecn_eligible && bbr_can_use_ecn(sk) && ++ bbr_param(sk, ecn_factor) && ++ (bbr->min_rtt_us <= bbr_ecn_max_rtt_us || ++ !bbr_ecn_max_rtt_us)) ++ bbr->ecn_eligible = 1; ++ ++ /* Skip updating alpha only if not ECN-eligible and PLB is disabled. */ ++ want_ecn_alpha = (bbr->ecn_eligible || ++ (bbr_can_use_ecn(sk) && ++ READ_ONCE(net->ipv4.sysctl_tcp_plb_enabled))); ++ if (!want_ecn_alpha) ++ return -1; ++ ++ delivered = tp->delivered - bbr->alpha_last_delivered; ++ delivered_ce = tp->delivered_ce - bbr->alpha_last_delivered_ce; ++ ++ if (delivered == 0 || /* avoid divide by zero */ ++ WARN_ON_ONCE(delivered < 0 || delivered_ce < 0)) /* backwards? */ ++ return -1; ++ ++ BUILD_BUG_ON(BBR_SCALE != TCP_PLB_SCALE); ++ ce_ratio = (u64)delivered_ce << BBR_SCALE; ++ do_div(ce_ratio, delivered); ++ ++ gain = bbr_param(sk, ecn_alpha_gain); ++ alpha = ((BBR_UNIT - gain) * bbr->ecn_alpha) >> BBR_SCALE; ++ alpha += (gain * ce_ratio) >> BBR_SCALE; ++ bbr->ecn_alpha = min_t(u32, alpha, BBR_UNIT); ++ ++ bbr->alpha_last_delivered = tp->delivered; ++ bbr->alpha_last_delivered_ce = tp->delivered_ce; ++ ++ bbr_check_ecn_too_high_in_startup(sk, ce_ratio); ++ return (int)ce_ratio; + } + +-/* In theory BBR does not need to undo the cwnd since it does not +- * always reduce cwnd on losses (see bbr_main()). Keep it for now. ++/* Protective Load Balancing (PLB). PLB rehashes outgoing data (to a new IPv6 ++ * flow label) if it encounters sustained congestion in the form of ECN marks. + */ +-__bpf_kfunc static u32 bbr_undo_cwnd(struct sock *sk) ++static void bbr_plb(struct sock *sk, const struct rate_sample *rs, int ce_ratio) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (bbr->round_start && ce_ratio >= 0) ++ tcp_plb_update_state(sk, &bbr->plb, ce_ratio); ++ ++ tcp_plb_check_rehash(sk, &bbr->plb); ++} ++ ++/* Each round trip of BBR_BW_PROBE_UP, double volume of probing data. */ ++static void bbr_raise_inflight_hi_slope(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 growth_this_round, cnt; ++ ++ /* Calculate "slope": packets S/Acked per inflight_hi increment. */ ++ growth_this_round = 1 << bbr->bw_probe_up_rounds; ++ bbr->bw_probe_up_rounds = min(bbr->bw_probe_up_rounds + 1, 30); ++ cnt = tcp_snd_cwnd(tp) / growth_this_round; ++ cnt = max(cnt, 1U); ++ bbr->bw_probe_up_cnt = cnt; ++} ++ ++/* In BBR_BW_PROBE_UP, not seeing high loss/ECN/queue, so raise inflight_hi. */ ++static void bbr_probe_inflight_hi_upward(struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 delta; ++ ++ if (!tp->is_cwnd_limited || tcp_snd_cwnd(tp) < bbr->inflight_hi) ++ return; /* not fully using inflight_hi, so don't grow it */ ++ ++ /* For each bw_probe_up_cnt packets ACKed, increase inflight_hi by 1. */ ++ bbr->bw_probe_up_acks += rs->acked_sacked; ++ if (bbr->bw_probe_up_acks >= bbr->bw_probe_up_cnt) { ++ delta = bbr->bw_probe_up_acks / bbr->bw_probe_up_cnt; ++ bbr->bw_probe_up_acks -= delta * bbr->bw_probe_up_cnt; ++ bbr->inflight_hi += delta; ++ bbr->try_fast_path = 0; /* Need to update cwnd */ ++ } ++ ++ if (bbr->round_start) ++ bbr_raise_inflight_hi_slope(sk); ++} ++ ++/* Does loss/ECN rate for this sample say inflight is "too high"? ++ * This is used by both the bbr_check_loss_too_high_in_startup() function, ++ * which can be used in either v1 or v2, and the PROBE_UP phase of v2, which ++ * uses it to notice when loss/ECN rates suggest inflight is too high. ++ */ ++static bool bbr_is_inflight_too_high(const struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ const struct bbr *bbr = inet_csk_ca(sk); ++ u32 loss_thresh, ecn_thresh; ++ ++ if (rs->lost > 0 && rs->tx_in_flight) { ++ loss_thresh = (u64)rs->tx_in_flight * bbr_param(sk, loss_thresh) >> ++ BBR_SCALE; ++ if (rs->lost > loss_thresh) { ++ return true; ++ } ++ } ++ ++ if (rs->delivered_ce > 0 && rs->delivered > 0 && ++ bbr->ecn_eligible && bbr_param(sk, ecn_thresh)) { ++ ecn_thresh = (u64)rs->delivered * bbr_param(sk, ecn_thresh) >> ++ BBR_SCALE; ++ if (rs->delivered_ce > ecn_thresh) { ++ return true; ++ } ++ } ++ ++ return false; ++} ++ ++/* Calculate the tx_in_flight level that corresponded to excessive loss. ++ * We find "lost_prefix" segs of the skb where loss rate went too high, ++ * by solving for "lost_prefix" in the following equation: ++ * lost / inflight >= loss_thresh ++ * (lost_prev + lost_prefix) / (inflight_prev + lost_prefix) >= loss_thresh ++ * Then we take that equation, convert it to fixed point, and ++ * round up to the nearest packet. ++ */ ++static u32 bbr_inflight_hi_from_lost_skb(const struct sock *sk, ++ const struct rate_sample *rs, ++ const struct sk_buff *skb) ++{ ++ const struct tcp_sock *tp = tcp_sk(sk); ++ u32 loss_thresh = bbr_param(sk, loss_thresh); ++ u32 pcount, divisor, inflight_hi; ++ s32 inflight_prev, lost_prev; ++ u64 loss_budget, lost_prefix; ++ ++ pcount = tcp_skb_pcount(skb); ++ ++ /* How much data was in flight before this skb? */ ++ inflight_prev = rs->tx_in_flight - pcount; ++ if (inflight_prev < 0) { ++ WARN_ONCE(tcp_skb_tx_in_flight_is_suspicious( ++ pcount, ++ TCP_SKB_CB(skb)->sacked, ++ rs->tx_in_flight), ++ "tx_in_flight: %u pcount: %u reneg: %u", ++ rs->tx_in_flight, pcount, tcp_sk(sk)->is_sack_reneg); ++ return ~0U; ++ } ++ ++ /* How much inflight data was marked lost before this skb? */ ++ lost_prev = rs->lost - pcount; ++ if (WARN_ONCE(lost_prev < 0, ++ "cwnd: %u ca: %d out: %u lost: %u pif: %u " ++ "tx_in_flight: %u tx.lost: %u tp->lost: %u rs->lost: %d " ++ "lost_prev: %d pcount: %d seq: %u end_seq: %u reneg: %u", ++ tcp_snd_cwnd(tp), inet_csk(sk)->icsk_ca_state, ++ tp->packets_out, tp->lost_out, tcp_packets_in_flight(tp), ++ rs->tx_in_flight, TCP_SKB_CB(skb)->tx.lost, tp->lost, ++ rs->lost, lost_prev, pcount, ++ TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq, ++ tp->is_sack_reneg)) ++ return ~0U; ++ ++ /* At what prefix of this lost skb did losss rate exceed loss_thresh? */ ++ loss_budget = (u64)inflight_prev * loss_thresh + BBR_UNIT - 1; ++ loss_budget >>= BBR_SCALE; ++ if (lost_prev >= loss_budget) { ++ lost_prefix = 0; /* previous losses crossed loss_thresh */ ++ } else { ++ lost_prefix = loss_budget - lost_prev; ++ lost_prefix <<= BBR_SCALE; ++ divisor = BBR_UNIT - loss_thresh; ++ if (WARN_ON_ONCE(!divisor)) /* loss_thresh is 8 bits */ ++ return ~0U; ++ do_div(lost_prefix, divisor); ++ } ++ ++ inflight_hi = inflight_prev + lost_prefix; ++ return inflight_hi; ++} ++ ++/* If loss/ECN rates during probing indicated we may have overfilled a ++ * buffer, return an operating point that tries to leave unutilized headroom in ++ * the path for other flows, for fairness convergence and lower RTTs and loss. ++ */ ++static u32 bbr_inflight_with_headroom(const struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 headroom, headroom_fraction; ++ ++ if (bbr->inflight_hi == ~0U) ++ return ~0U; ++ ++ headroom_fraction = bbr_param(sk, inflight_headroom); ++ headroom = ((u64)bbr->inflight_hi * headroom_fraction) >> BBR_SCALE; ++ headroom = max(headroom, 1U); ++ return max_t(s32, bbr->inflight_hi - headroom, ++ bbr_param(sk, cwnd_min_target)); ++} ++ ++/* Bound cwnd to a sensible level, based on our current probing state ++ * machine phase and model of a good inflight level (inflight_lo, inflight_hi). ++ */ ++static void bbr_bound_cwnd_for_inflight_model(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 cap; ++ ++ /* tcp_rcv_synsent_state_process() currently calls tcp_ack() ++ * and thus cong_control() without first initializing us(!). ++ */ ++ if (!bbr->initialized) ++ return; ++ ++ cap = ~0U; ++ if (bbr->mode == BBR_PROBE_BW && ++ bbr->cycle_idx != BBR_BW_PROBE_CRUISE) { ++ /* Probe to see if more packets fit in the path. */ ++ cap = bbr->inflight_hi; ++ } else { ++ if (bbr->mode == BBR_PROBE_RTT || ++ (bbr->mode == BBR_PROBE_BW && ++ bbr->cycle_idx == BBR_BW_PROBE_CRUISE)) ++ cap = bbr_inflight_with_headroom(sk); ++ } ++ /* Adapt to any loss/ECN since our last bw probe. */ ++ cap = min(cap, bbr->inflight_lo); ++ ++ cap = max_t(u32, cap, bbr_param(sk, cwnd_min_target)); ++ tcp_snd_cwnd_set(tp, min(cap, tcp_snd_cwnd(tp))); ++} ++ ++/* How should we multiplicatively cut bw or inflight limits based on ECN? */ ++static u32 bbr_ecn_cut(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ return BBR_UNIT - ++ ((bbr->ecn_alpha * bbr_param(sk, ecn_factor)) >> BBR_SCALE); ++} ++ ++/* Init lower bounds if have not inited yet. */ ++static void bbr_init_lower_bounds(struct sock *sk, bool init_bw) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (init_bw && bbr->bw_lo == ~0U) ++ bbr->bw_lo = bbr_max_bw(sk); ++ if (bbr->inflight_lo == ~0U) ++ bbr->inflight_lo = tcp_snd_cwnd(tp); ++} ++ ++/* Reduce bw and inflight to (1 - beta). */ ++static void bbr_loss_lower_bounds(struct sock *sk, u32 *bw, u32 *inflight) ++{ ++ struct bbr* bbr = inet_csk_ca(sk); ++ u32 loss_cut = BBR_UNIT - bbr_param(sk, beta); ++ ++ *bw = max_t(u32, bbr->bw_latest, ++ (u64)bbr->bw_lo * loss_cut >> BBR_SCALE); ++ *inflight = max_t(u32, bbr->inflight_latest, ++ (u64)bbr->inflight_lo * loss_cut >> BBR_SCALE); ++} ++ ++/* Reduce inflight to (1 - alpha*ecn_factor). */ ++static void bbr_ecn_lower_bounds(struct sock *sk, u32 *inflight) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 ecn_cut = bbr_ecn_cut(sk); ++ ++ *inflight = (u64)bbr->inflight_lo * ecn_cut >> BBR_SCALE; ++} ++ ++/* Estimate a short-term lower bound on the capacity available now, based ++ * on measurements of the current delivery process and recent history. When we ++ * are seeing loss/ECN at times when we are not probing bw, then conservatively ++ * move toward flow balance by multiplicatively cutting our short-term ++ * estimated safe rate and volume of data (bw_lo and inflight_lo). We use a ++ * multiplicative decrease in order to converge to a lower capacity in time ++ * logarithmic in the magnitude of the decrease. ++ * ++ * However, we do not cut our short-term estimates lower than the current rate ++ * and volume of delivered data from this round trip, since from the current ++ * delivery process we can estimate the measured capacity available now. ++ * ++ * Anything faster than that approach would knowingly risk high loss, which can ++ * cause low bw for Reno/CUBIC and high loss recovery latency for ++ * request/response flows using any congestion control. ++ */ ++static void bbr_adapt_lower_bounds(struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 ecn_inflight_lo = ~0U; ++ ++ /* We only use lower-bound estimates when not probing bw. ++ * When probing we need to push inflight higher to probe bw. ++ */ ++ if (bbr_is_probing_bandwidth(sk)) ++ return; ++ ++ /* ECN response. */ ++ if (bbr->ecn_in_round && bbr_param(sk, ecn_factor)) { ++ bbr_init_lower_bounds(sk, false); ++ bbr_ecn_lower_bounds(sk, &ecn_inflight_lo); ++ } ++ ++ /* Loss response. */ ++ if (bbr->loss_in_round) { ++ bbr_init_lower_bounds(sk, true); ++ bbr_loss_lower_bounds(sk, &bbr->bw_lo, &bbr->inflight_lo); ++ } ++ ++ /* Adjust to the lower of the levels implied by loss/ECN. */ ++ bbr->inflight_lo = min(bbr->inflight_lo, ecn_inflight_lo); ++ bbr->bw_lo = max(1U, bbr->bw_lo); ++} ++ ++/* Reset any short-term lower-bound adaptation to congestion, so that we can ++ * push our inflight up. ++ */ ++static void bbr_reset_lower_bounds(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->bw_lo = ~0U; ++ bbr->inflight_lo = ~0U; ++} ++ ++/* After bw probing (STARTUP/PROBE_UP), reset signals before entering a state ++ * machine phase where we adapt our lower bound based on congestion signals. ++ */ ++static void bbr_reset_congestion_signals(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->loss_in_round = 0; ++ bbr->ecn_in_round = 0; ++ bbr->loss_in_cycle = 0; ++ bbr->ecn_in_cycle = 0; ++ bbr->bw_latest = 0; ++ bbr->inflight_latest = 0; ++} ++ ++static void bbr_exit_loss_recovery(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ tcp_snd_cwnd_set(tp, max(tcp_snd_cwnd(tp), bbr->prior_cwnd)); ++ bbr->try_fast_path = 0; /* bound cwnd using latest model */ ++} ++ ++/* Update rate and volume of delivered data from latest round trip. */ ++static void bbr_update_latest_delivery_signals( ++ struct sock *sk, const struct rate_sample *rs, struct bbr_context *ctx) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->loss_round_start = 0; ++ if (rs->interval_us <= 0 || !rs->acked_sacked) ++ return; /* Not a valid observation */ ++ ++ bbr->bw_latest = max_t(u32, bbr->bw_latest, ctx->sample_bw); ++ bbr->inflight_latest = max_t(u32, bbr->inflight_latest, rs->delivered); ++ ++ if (!before(rs->prior_delivered, bbr->loss_round_delivered)) { ++ bbr->loss_round_delivered = tp->delivered; ++ bbr->loss_round_start = 1; /* mark start of new round trip */ ++ } ++} ++ ++/* Once per round, reset filter for latest rate and volume of delivered data. */ ++static void bbr_advance_latest_delivery_signals( ++ struct sock *sk, const struct rate_sample *rs, struct bbr_context *ctx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ /* If ACK matches a TLP retransmit, persist the filter. If we detect ++ * that a TLP retransmit plugged a tail loss, we'll want to remember ++ * how much data the path delivered before the tail loss. ++ */ ++ if (bbr->loss_round_start && !rs->is_acking_tlp_retrans_seq) { ++ bbr->bw_latest = ctx->sample_bw; ++ bbr->inflight_latest = rs->delivered; ++ } ++} ++ ++/* Update (most of) our congestion signals: track the recent rate and volume of ++ * delivered data, presence of loss, and EWMA degree of ECN marking. ++ */ ++static void bbr_update_congestion_signals( ++ struct sock *sk, const struct rate_sample *rs, struct bbr_context *ctx) + { + struct bbr *bbr = inet_csk_ca(sk); ++ u64 bw; ++ ++ if (rs->interval_us <= 0 || !rs->acked_sacked) ++ return; /* Not a valid observation */ ++ bw = ctx->sample_bw; + +- bbr->full_bw = 0; /* spurious slow-down; reset full pipe detection */ ++ if (!rs->is_app_limited || bw >= bbr_max_bw(sk)) ++ bbr_take_max_bw_sample(sk, bw); ++ ++ bbr->loss_in_round |= (rs->losses > 0); ++ ++ if (!bbr->loss_round_start) ++ return; /* skip the per-round-trip updates */ ++ /* Now do per-round-trip updates. */ ++ bbr_adapt_lower_bounds(sk, rs); ++ ++ bbr->loss_in_round = 0; ++ bbr->ecn_in_round = 0; ++} ++ ++/* Bandwidth probing can cause loss. To help coexistence with loss-based ++ * congestion control we spread out our probing in a Reno-conscious way. Due to ++ * the shape of the Reno sawtooth, the time required between loss epochs for an ++ * idealized Reno flow is a number of round trips that is the BDP of that ++ * flow. We count packet-timed round trips directly, since measured RTT can ++ * vary widely, and Reno is driven by packet-timed round trips. ++ */ ++static bool bbr_is_reno_coexistence_probe_time(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 rounds; ++ ++ /* Random loss can shave some small percentage off of our inflight ++ * in each round. To survive this, flows need robust periodic probes. ++ */ ++ rounds = min_t(u32, bbr_param(sk, bw_probe_max_rounds), bbr_target_inflight(sk)); ++ return bbr->rounds_since_probe >= rounds; ++} ++ ++/* How long do we want to wait before probing for bandwidth (and risking ++ * loss)? We randomize the wait, for better mixing and fairness convergence. ++ * ++ * We bound the Reno-coexistence inter-bw-probe time to be 62-63 round trips. ++ * This is calculated to allow fairness with a 25Mbps, 30ms Reno flow, ++ * (eg 4K video to a broadband user): ++ * BDP = 25Mbps * .030sec /(1514bytes) = 61.9 packets ++ * ++ * We bound the BBR-native inter-bw-probe wall clock time to be: ++ * (a) higher than 2 sec: to try to avoid causing loss for a long enough time ++ * to allow Reno at 30ms to get 4K video bw, the inter-bw-probe time must ++ * be at least: 25Mbps * .030sec / (1514bytes) * 0.030sec = 1.9secs ++ * (b) lower than 3 sec: to ensure flows can start probing in a reasonable ++ * amount of time to discover unutilized bw on human-scale interactive ++ * time-scales (e.g. perhaps traffic from a web page download that we ++ * were competing with is now complete). ++ */ ++static void bbr_pick_probe_wait(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ /* Decide the random round-trip bound for wait until probe: */ ++ bbr->rounds_since_probe = ++ get_random_u32_below(bbr_param(sk, bw_probe_rand_rounds)); ++ /* Decide the random wall clock bound for wait until probe: */ ++ bbr->probe_wait_us = bbr_param(sk, bw_probe_base_us) + ++ get_random_u32_below(bbr_param(sk, bw_probe_rand_us)); ++} ++ ++static void bbr_set_cycle_idx(struct sock *sk, int cycle_idx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->cycle_idx = cycle_idx; ++ /* New phase, so need to update cwnd and pacing rate. */ ++ bbr->try_fast_path = 0; ++} ++ ++/* Send at estimated bw to fill the pipe, but not queue. We need this phase ++ * before PROBE_UP, because as soon as we send faster than the available bw ++ * we will start building a queue, and if the buffer is shallow we can cause ++ * loss. If we do not fill the pipe before we cause this loss, our bw_hi and ++ * inflight_hi estimates will underestimate. ++ */ ++static void bbr_start_bw_probe_refill(struct sock *sk, u32 bw_probe_up_rounds) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr_reset_lower_bounds(sk); ++ bbr->bw_probe_up_rounds = bw_probe_up_rounds; ++ bbr->bw_probe_up_acks = 0; ++ bbr->stopped_risky_probe = 0; ++ bbr->ack_phase = BBR_ACKS_REFILLING; ++ bbr->next_rtt_delivered = tp->delivered; ++ bbr_set_cycle_idx(sk, BBR_BW_PROBE_REFILL); ++} ++ ++/* Now probe max deliverable data rate and volume. */ ++static void bbr_start_bw_probe_up(struct sock *sk, struct bbr_context *ctx) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->ack_phase = BBR_ACKS_PROBE_STARTING; ++ bbr->next_rtt_delivered = tp->delivered; ++ bbr->cycle_mstamp = tp->tcp_mstamp; ++ bbr_reset_full_bw(sk); ++ bbr->full_bw = ctx->sample_bw; ++ bbr_set_cycle_idx(sk, BBR_BW_PROBE_UP); ++ bbr_raise_inflight_hi_slope(sk); ++} ++ ++/* Start a new PROBE_BW probing cycle of some wall clock length. Pick a wall ++ * clock time at which to probe beyond an inflight that we think to be ++ * safe. This will knowingly risk packet loss, so we want to do this rarely, to ++ * keep packet loss rates low. Also start a round-trip counter, to probe faster ++ * if we estimate a Reno flow at our BDP would probe faster. ++ */ ++static void bbr_start_bw_probe_down(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr_reset_congestion_signals(sk); ++ bbr->bw_probe_up_cnt = ~0U; /* not growing inflight_hi any more */ ++ bbr_pick_probe_wait(sk); ++ bbr->cycle_mstamp = tp->tcp_mstamp; /* start wall clock */ ++ bbr->ack_phase = BBR_ACKS_PROBE_STOPPING; ++ bbr->next_rtt_delivered = tp->delivered; ++ bbr_set_cycle_idx(sk, BBR_BW_PROBE_DOWN); ++} ++ ++/* Cruise: maintain what we estimate to be a neutral, conservative ++ * operating point, without attempting to probe up for bandwidth or down for ++ * RTT, and only reducing inflight in response to loss/ECN signals. ++ */ ++static void bbr_start_bw_probe_cruise(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (bbr->inflight_lo != ~0U) ++ bbr->inflight_lo = min(bbr->inflight_lo, bbr->inflight_hi); ++ ++ bbr_set_cycle_idx(sk, BBR_BW_PROBE_CRUISE); ++} ++ ++/* Loss and/or ECN rate is too high while probing. ++ * Adapt (once per bw probe) by cutting inflight_hi and then restarting cycle. ++ */ ++static void bbr_handle_inflight_too_high(struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ const u32 beta = bbr_param(sk, beta); ++ ++ bbr->prev_probe_too_high = 1; ++ bbr->bw_probe_samples = 0; /* only react once per probe */ ++ /* If we are app-limited then we are not robustly ++ * probing the max volume of inflight data we think ++ * might be safe (analogous to how app-limited bw ++ * samples are not known to be robustly probing bw). ++ */ ++ if (!rs->is_app_limited) { ++ bbr->inflight_hi = max_t(u32, rs->tx_in_flight, ++ (u64)bbr_target_inflight(sk) * ++ (BBR_UNIT - beta) >> BBR_SCALE); ++ } ++ if (bbr->mode == BBR_PROBE_BW && bbr->cycle_idx == BBR_BW_PROBE_UP) ++ bbr_start_bw_probe_down(sk); ++} ++ ++/* If we're seeing bw and loss samples reflecting our bw probing, adapt ++ * using the signals we see. If loss or ECN mark rate gets too high, then adapt ++ * inflight_hi downward. If we're able to push inflight higher without such ++ * signals, push higher: adapt inflight_hi upward. ++ */ ++static bool bbr_adapt_upper_bounds(struct sock *sk, ++ const struct rate_sample *rs, ++ struct bbr_context *ctx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ /* Track when we'll see bw/loss samples resulting from our bw probes. */ ++ if (bbr->ack_phase == BBR_ACKS_PROBE_STARTING && bbr->round_start) ++ bbr->ack_phase = BBR_ACKS_PROBE_FEEDBACK; ++ if (bbr->ack_phase == BBR_ACKS_PROBE_STOPPING && bbr->round_start) { ++ /* End of samples from bw probing phase. */ ++ bbr->bw_probe_samples = 0; ++ bbr->ack_phase = BBR_ACKS_INIT; ++ /* At this point in the cycle, our current bw sample is also ++ * our best recent chance at finding the highest available bw ++ * for this flow. So now is the best time to forget the bw ++ * samples from the previous cycle, by advancing the window. ++ */ ++ if (bbr->mode == BBR_PROBE_BW && !rs->is_app_limited) ++ bbr_advance_max_bw_filter(sk); ++ /* If we had an inflight_hi, then probed and pushed inflight all ++ * the way up to hit that inflight_hi without seeing any ++ * high loss/ECN in all the resulting ACKs from that probing, ++ * then probe up again, this time letting inflight persist at ++ * inflight_hi for a round trip, then accelerating beyond. ++ */ ++ if (bbr->mode == BBR_PROBE_BW && ++ bbr->stopped_risky_probe && !bbr->prev_probe_too_high) { ++ bbr_start_bw_probe_refill(sk, 0); ++ return true; /* yes, decided state transition */ ++ } ++ } ++ if (bbr_is_inflight_too_high(sk, rs)) { ++ if (bbr->bw_probe_samples) /* sample is from bw probing? */ ++ bbr_handle_inflight_too_high(sk, rs); ++ } else { ++ /* Loss/ECN rate is declared safe. Adjust upper bound upward. */ ++ ++ if (bbr->inflight_hi == ~0U) ++ return false; /* no excess queue signals yet */ ++ ++ /* To be resilient to random loss, we must raise bw/inflight_hi ++ * if we observe in any phase that a higher level is safe. ++ */ ++ if (rs->tx_in_flight > bbr->inflight_hi) { ++ bbr->inflight_hi = rs->tx_in_flight; ++ } ++ ++ if (bbr->mode == BBR_PROBE_BW && ++ bbr->cycle_idx == BBR_BW_PROBE_UP) ++ bbr_probe_inflight_hi_upward(sk, rs); ++ } ++ ++ return false; ++} ++ ++/* Check if it's time to probe for bandwidth now, and if so, kick it off. */ ++static bool bbr_check_time_to_probe_bw(struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 n; ++ ++ /* If we seem to be at an operating point where we are not seeing loss ++ * but we are seeing ECN marks, then when the ECN marks cease we reprobe ++ * quickly (in case cross-traffic has ceased and freed up bw). ++ */ ++ if (bbr_param(sk, ecn_reprobe_gain) && bbr->ecn_eligible && ++ bbr->ecn_in_cycle && !bbr->loss_in_cycle && ++ inet_csk(sk)->icsk_ca_state == TCP_CA_Open) { ++ /* Calculate n so that when bbr_raise_inflight_hi_slope() ++ * computes growth_this_round as 2^n it will be roughly the ++ * desired volume of data (inflight_hi*ecn_reprobe_gain). ++ */ ++ n = ilog2((((u64)bbr->inflight_hi * ++ bbr_param(sk, ecn_reprobe_gain)) >> BBR_SCALE)); ++ bbr_start_bw_probe_refill(sk, n); ++ return true; ++ } ++ ++ if (bbr_has_elapsed_in_phase(sk, bbr->probe_wait_us) || ++ bbr_is_reno_coexistence_probe_time(sk)) { ++ bbr_start_bw_probe_refill(sk, 0); ++ return true; ++ } ++ return false; ++} ++ ++/* Is it time to transition from PROBE_DOWN to PROBE_CRUISE? */ ++static bool bbr_check_time_to_cruise(struct sock *sk, u32 inflight, u32 bw) ++{ ++ /* Always need to pull inflight down to leave headroom in queue. */ ++ if (inflight > bbr_inflight_with_headroom(sk)) ++ return false; ++ ++ return inflight <= bbr_inflight(sk, bw, BBR_UNIT); ++} ++ ++/* PROBE_BW state machine: cruise, refill, probe for bw, or drain? */ ++static void bbr_update_cycle_phase(struct sock *sk, ++ const struct rate_sample *rs, ++ struct bbr_context *ctx) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ bool is_bw_probe_done = false; ++ u32 inflight, bw; ++ ++ if (!bbr_full_bw_reached(sk)) ++ return; ++ ++ /* In DRAIN, PROBE_BW, or PROBE_RTT, adjust upper bounds. */ ++ if (bbr_adapt_upper_bounds(sk, rs, ctx)) ++ return; /* already decided state transition */ ++ ++ if (bbr->mode != BBR_PROBE_BW) ++ return; ++ ++ inflight = bbr_packets_in_net_at_edt(sk, rs->prior_in_flight); ++ bw = bbr_max_bw(sk); ++ ++ switch (bbr->cycle_idx) { ++ /* First we spend most of our time cruising with a pacing_gain of 1.0, ++ * which paces at the estimated bw, to try to fully use the pipe ++ * without building queue. If we encounter loss/ECN marks, we adapt ++ * by slowing down. ++ */ ++ case BBR_BW_PROBE_CRUISE: ++ if (bbr_check_time_to_probe_bw(sk, rs)) ++ return; /* already decided state transition */ ++ break; ++ ++ /* After cruising, when it's time to probe, we first "refill": we send ++ * at the estimated bw to fill the pipe, before probing higher and ++ * knowingly risking overflowing the bottleneck buffer (causing loss). ++ */ ++ case BBR_BW_PROBE_REFILL: ++ if (bbr->round_start) { ++ /* After one full round trip of sending in REFILL, we ++ * start to see bw samples reflecting our REFILL, which ++ * may be putting too much data in flight. ++ */ ++ bbr->bw_probe_samples = 1; ++ bbr_start_bw_probe_up(sk, ctx); ++ } ++ break; ++ ++ /* After we refill the pipe, we probe by using a pacing_gain > 1.0, to ++ * probe for bw. If we have not seen loss/ECN, we try to raise inflight ++ * to at least pacing_gain*BDP; note that this may take more than ++ * min_rtt if min_rtt is small (e.g. on a LAN). ++ * ++ * We terminate PROBE_UP bandwidth probing upon any of the following: ++ * ++ * (1) We've pushed inflight up to hit the inflight_hi target set in the ++ * most recent previous bw probe phase. Thus we want to start ++ * draining the queue immediately because it's very likely the most ++ * recently sent packets will fill the queue and cause drops. ++ * (2) If inflight_hi has not limited bandwidth growth recently, and ++ * yet delivered bandwidth has not increased much recently ++ * (bbr->full_bw_now). ++ * (3) Loss filter says loss rate is "too high". ++ * (4) ECN filter says ECN mark rate is "too high". ++ * ++ * (1) (2) checked here, (3) (4) checked in bbr_is_inflight_too_high() ++ */ ++ case BBR_BW_PROBE_UP: ++ if (bbr->prev_probe_too_high && ++ inflight >= bbr->inflight_hi) { ++ bbr->stopped_risky_probe = 1; ++ is_bw_probe_done = true; ++ } else { ++ if (tp->is_cwnd_limited && ++ tcp_snd_cwnd(tp) >= bbr->inflight_hi) { ++ /* inflight_hi is limiting bw growth */ ++ bbr_reset_full_bw(sk); ++ bbr->full_bw = ctx->sample_bw; ++ } else if (bbr->full_bw_now) { ++ /* Plateau in estimated bw. Pipe looks full. */ ++ is_bw_probe_done = true; ++ } ++ } ++ if (is_bw_probe_done) { ++ bbr->prev_probe_too_high = 0; /* no loss/ECN (yet) */ ++ bbr_start_bw_probe_down(sk); /* restart w/ down */ ++ } ++ break; ++ ++ /* After probing in PROBE_UP, we have usually accumulated some data in ++ * the bottleneck buffer (if bw probing didn't find more bw). We next ++ * enter PROBE_DOWN to try to drain any excess data from the queue. To ++ * do this, we use a pacing_gain < 1.0. We hold this pacing gain until ++ * our inflight is less then that target cruising point, which is the ++ * minimum of (a) the amount needed to leave headroom, and (b) the ++ * estimated BDP. Once inflight falls to match the target, we estimate ++ * the queue is drained; persisting would underutilize the pipe. ++ */ ++ case BBR_BW_PROBE_DOWN: ++ if (bbr_check_time_to_probe_bw(sk, rs)) ++ return; /* already decided state transition */ ++ if (bbr_check_time_to_cruise(sk, inflight, bw)) ++ bbr_start_bw_probe_cruise(sk); ++ break; ++ ++ default: ++ WARN_ONCE(1, "BBR invalid cycle index %u\n", bbr->cycle_idx); ++ } ++} ++ ++/* Exiting PROBE_RTT, so return to bandwidth probing in STARTUP or PROBE_BW. */ ++static void bbr_exit_probe_rtt(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr_reset_lower_bounds(sk); ++ if (bbr_full_bw_reached(sk)) { ++ bbr->mode = BBR_PROBE_BW; ++ /* Raising inflight after PROBE_RTT may cause loss, so reset ++ * the PROBE_BW clock and schedule the next bandwidth probe for ++ * a friendly and randomized future point in time. ++ */ ++ bbr_start_bw_probe_down(sk); ++ /* Since we are exiting PROBE_RTT, we know inflight is ++ * below our estimated BDP, so it is reasonable to cruise. ++ */ ++ bbr_start_bw_probe_cruise(sk); ++ } else { ++ bbr->mode = BBR_STARTUP; ++ } ++} ++ ++/* Exit STARTUP based on loss rate > 1% and loss gaps in round >= N. Wait until ++ * the end of the round in recovery to get a good estimate of how many packets ++ * have been lost, and how many we need to drain with a low pacing rate. ++ */ ++static void bbr_check_loss_too_high_in_startup(struct sock *sk, ++ const struct rate_sample *rs) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (bbr_full_bw_reached(sk)) ++ return; ++ ++ /* For STARTUP exit, check the loss rate at the end of each round trip ++ * of Recovery episodes in STARTUP. We check the loss rate at the end ++ * of the round trip to filter out noisy/low loss and have a better ++ * sense of inflight (extent of loss), so we can drain more accurately. ++ */ ++ if (rs->losses && bbr->loss_events_in_round < 0xf) ++ bbr->loss_events_in_round++; /* update saturating counter */ ++ if (bbr_param(sk, full_loss_cnt) && bbr->loss_round_start && ++ inet_csk(sk)->icsk_ca_state == TCP_CA_Recovery && ++ bbr->loss_events_in_round >= bbr_param(sk, full_loss_cnt) && ++ bbr_is_inflight_too_high(sk, rs)) { ++ bbr_handle_queue_too_high_in_startup(sk); ++ return; ++ } ++ if (bbr->loss_round_start) ++ bbr->loss_events_in_round = 0; ++} ++ ++/* Estimate when the pipe is full, using the change in delivery rate: BBR ++ * estimates bw probing filled the pipe if the estimated bw hasn't changed by ++ * at least bbr_full_bw_thresh (25%) after bbr_full_bw_cnt (3) non-app-limited ++ * rounds. Why 3 rounds: 1: rwin autotuning grows the rwin, 2: we fill the ++ * higher rwin, 3: we get higher delivery rate samples. Or transient ++ * cross-traffic or radio noise can go away. CUBIC Hystart shares a similar ++ * design goal, but uses delay and inter-ACK spacing instead of bandwidth. ++ */ ++static void bbr_check_full_bw_reached(struct sock *sk, ++ const struct rate_sample *rs, ++ struct bbr_context *ctx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 bw_thresh, full_cnt, thresh; ++ ++ if (bbr->full_bw_now || rs->is_app_limited) ++ return; ++ ++ thresh = bbr_param(sk, full_bw_thresh); ++ full_cnt = bbr_param(sk, full_bw_cnt); ++ bw_thresh = (u64)bbr->full_bw * thresh >> BBR_SCALE; ++ if (ctx->sample_bw >= bw_thresh) { ++ bbr_reset_full_bw(sk); ++ bbr->full_bw = ctx->sample_bw; ++ return; ++ } ++ if (!bbr->round_start) ++ return; ++ ++bbr->full_bw_cnt; ++ bbr->full_bw_now = bbr->full_bw_cnt >= full_cnt; ++ bbr->full_bw_reached |= bbr->full_bw_now; ++} ++ ++/* If pipe is probably full, drain the queue and then enter steady-state. */ ++static void bbr_check_drain(struct sock *sk, const struct rate_sample *rs, ++ struct bbr_context *ctx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ if (bbr->mode == BBR_STARTUP && bbr_full_bw_reached(sk)) { ++ bbr->mode = BBR_DRAIN; /* drain queue we created */ ++ /* Set ssthresh to export purely for monitoring, to signal ++ * completion of initial STARTUP by setting to a non- ++ * TCP_INFINITE_SSTHRESH value (ssthresh is not used by BBR). ++ */ ++ tcp_sk(sk)->snd_ssthresh = ++ bbr_inflight(sk, bbr_max_bw(sk), BBR_UNIT); ++ bbr_reset_congestion_signals(sk); ++ } /* fall through to check if in-flight is already small: */ ++ if (bbr->mode == BBR_DRAIN && ++ bbr_packets_in_net_at_edt(sk, tcp_packets_in_flight(tcp_sk(sk))) <= ++ bbr_inflight(sk, bbr_max_bw(sk), BBR_UNIT)) { ++ bbr->mode = BBR_PROBE_BW; ++ bbr_start_bw_probe_down(sk); ++ } ++} ++ ++static void bbr_update_model(struct sock *sk, const struct rate_sample *rs, ++ struct bbr_context *ctx) ++{ ++ bbr_update_congestion_signals(sk, rs, ctx); ++ bbr_update_ack_aggregation(sk, rs); ++ bbr_check_loss_too_high_in_startup(sk, rs); ++ bbr_check_full_bw_reached(sk, rs, ctx); ++ bbr_check_drain(sk, rs, ctx); ++ bbr_update_cycle_phase(sk, rs, ctx); ++ bbr_update_min_rtt(sk, rs); ++} ++ ++/* Fast path for app-limited case. ++ * ++ * On each ack, we execute bbr state machine, which primarily consists of: ++ * 1) update model based on new rate sample, and ++ * 2) update control based on updated model or state change. ++ * ++ * There are certain workload/scenarios, e.g. app-limited case, where ++ * either we can skip updating model or we can skip update of both model ++ * as well as control. This provides signifcant softirq cpu savings for ++ * processing incoming acks. ++ * ++ * In case of app-limited, if there is no congestion (loss/ecn) and ++ * if observed bw sample is less than current estimated bw, then we can ++ * skip some of the computation in bbr state processing: ++ * ++ * - if there is no rtt/mode/phase change: In this case, since all the ++ * parameters of the network model are constant, we can skip model ++ * as well control update. ++ * ++ * - else we can skip rest of the model update. But we still need to ++ * update the control to account for the new rtt/mode/phase. ++ * ++ * Returns whether we can take fast path or not. ++ */ ++static bool bbr_run_fast_path(struct sock *sk, bool *update_model, ++ const struct rate_sample *rs, struct bbr_context *ctx) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ u32 prev_min_rtt_us, prev_mode; ++ ++ if (bbr_param(sk, fast_path) && bbr->try_fast_path && ++ rs->is_app_limited && ctx->sample_bw < bbr_max_bw(sk) && ++ !bbr->loss_in_round && !bbr->ecn_in_round ) { ++ prev_mode = bbr->mode; ++ prev_min_rtt_us = bbr->min_rtt_us; ++ bbr_check_drain(sk, rs, ctx); ++ bbr_update_cycle_phase(sk, rs, ctx); ++ bbr_update_min_rtt(sk, rs); ++ ++ if (bbr->mode == prev_mode && ++ bbr->min_rtt_us == prev_min_rtt_us && ++ bbr->try_fast_path) { ++ return true; ++ } ++ ++ /* Skip model update, but control still needs to be updated */ ++ *update_model = false; ++ } ++ return false; ++} ++ ++__bpf_kfunc static void bbr_main(struct sock *sk, u32 ack, int flag, const struct rate_sample *rs) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ struct bbr_context ctx = { 0 }; ++ bool update_model = true; ++ u32 bw, round_delivered; ++ int ce_ratio = -1; ++ ++ round_delivered = bbr_update_round_start(sk, rs, &ctx); ++ if (bbr->round_start) { ++ bbr->rounds_since_probe = ++ min_t(s32, bbr->rounds_since_probe + 1, 0xFF); ++ ce_ratio = bbr_update_ecn_alpha(sk); ++ } ++ bbr_plb(sk, rs, ce_ratio); ++ ++ bbr->ecn_in_round |= (bbr->ecn_eligible && rs->is_ece); ++ bbr_calculate_bw_sample(sk, rs, &ctx); ++ bbr_update_latest_delivery_signals(sk, rs, &ctx); ++ ++ if (bbr_run_fast_path(sk, &update_model, rs, &ctx)) ++ goto out; ++ ++ if (update_model) ++ bbr_update_model(sk, rs, &ctx); ++ ++ bbr_update_gains(sk); ++ bw = bbr_bw(sk); ++ bbr_set_pacing_rate(sk, bw, bbr->pacing_gain); ++ bbr_set_cwnd(sk, rs, rs->acked_sacked, bw, bbr->cwnd_gain, ++ tcp_snd_cwnd(tp), &ctx); ++ bbr_bound_cwnd_for_inflight_model(sk); ++ ++out: ++ bbr_advance_latest_delivery_signals(sk, rs, &ctx); ++ bbr->prev_ca_state = inet_csk(sk)->icsk_ca_state; ++ bbr->loss_in_cycle |= rs->lost > 0; ++ bbr->ecn_in_cycle |= rs->delivered_ce > 0; ++} ++ ++__bpf_kfunc static void bbr_init(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr->initialized = 1; ++ ++ bbr->init_cwnd = min(0x7FU, tcp_snd_cwnd(tp)); ++ bbr->prior_cwnd = tp->prior_cwnd; ++ tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; ++ bbr->next_rtt_delivered = tp->delivered; ++ bbr->prev_ca_state = TCP_CA_Open; ++ ++ bbr->probe_rtt_done_stamp = 0; ++ bbr->probe_rtt_round_done = 0; ++ bbr->probe_rtt_min_us = tcp_min_rtt(tp); ++ bbr->probe_rtt_min_stamp = tcp_jiffies32; ++ bbr->min_rtt_us = tcp_min_rtt(tp); ++ bbr->min_rtt_stamp = tcp_jiffies32; ++ ++ bbr->has_seen_rtt = 0; ++ bbr_init_pacing_rate_from_rtt(sk); ++ ++ bbr->round_start = 0; ++ bbr->idle_restart = 0; ++ bbr->full_bw_reached = 0; ++ bbr->full_bw = 0; + bbr->full_bw_cnt = 0; +- bbr_reset_lt_bw_sampling(sk); +- return tcp_snd_cwnd(tcp_sk(sk)); ++ bbr->cycle_mstamp = 0; ++ bbr->cycle_idx = 0; ++ ++ bbr_reset_startup_mode(sk); ++ ++ bbr->ack_epoch_mstamp = tp->tcp_mstamp; ++ bbr->ack_epoch_acked = 0; ++ bbr->extra_acked_win_rtts = 0; ++ bbr->extra_acked_win_idx = 0; ++ bbr->extra_acked[0] = 0; ++ bbr->extra_acked[1] = 0; ++ ++ bbr->ce_state = 0; ++ bbr->prior_rcv_nxt = tp->rcv_nxt; ++ bbr->try_fast_path = 0; ++ ++ cmpxchg(&sk->sk_pacing_status, SK_PACING_NONE, SK_PACING_NEEDED); ++ ++ /* Start sampling ECN mark rate after first full flight is ACKed: */ ++ bbr->loss_round_delivered = tp->delivered + 1; ++ bbr->loss_round_start = 0; ++ bbr->undo_bw_lo = 0; ++ bbr->undo_inflight_lo = 0; ++ bbr->undo_inflight_hi = 0; ++ bbr->loss_events_in_round = 0; ++ bbr->startup_ecn_rounds = 0; ++ bbr_reset_congestion_signals(sk); ++ bbr->bw_lo = ~0U; ++ bbr->bw_hi[0] = 0; ++ bbr->bw_hi[1] = 0; ++ bbr->inflight_lo = ~0U; ++ bbr->inflight_hi = ~0U; ++ bbr_reset_full_bw(sk); ++ bbr->bw_probe_up_cnt = ~0U; ++ bbr->bw_probe_up_acks = 0; ++ bbr->bw_probe_up_rounds = 0; ++ bbr->probe_wait_us = 0; ++ bbr->stopped_risky_probe = 0; ++ bbr->ack_phase = BBR_ACKS_INIT; ++ bbr->rounds_since_probe = 0; ++ bbr->bw_probe_samples = 0; ++ bbr->prev_probe_too_high = 0; ++ bbr->ecn_eligible = 0; ++ bbr->ecn_alpha = bbr_param(sk, ecn_alpha_init); ++ bbr->alpha_last_delivered = 0; ++ bbr->alpha_last_delivered_ce = 0; ++ bbr->plb.pause_until = 0; ++ ++ tp->fast_ack_mode = bbr_fast_ack_mode ? 1 : 0; ++ ++ if (bbr_can_use_ecn(sk)) ++ tp->ecn_flags |= TCP_ECN_ECT_PERMANENT; ++} ++ ++/* BBR marks the current round trip as a loss round. */ ++static void bbr_note_loss(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ /* Capture "current" data over the full round trip of loss, to ++ * have a better chance of observing the full capacity of the path. ++ */ ++ if (!bbr->loss_in_round) /* first loss in this round trip? */ ++ bbr->loss_round_delivered = tp->delivered; /* set round trip */ ++ bbr->loss_in_round = 1; ++ bbr->loss_in_cycle = 1; + } + +-/* Entering loss recovery, so save cwnd for when we exit or undo recovery. */ ++/* Core TCP stack informs us that the given skb was just marked lost. */ ++__bpf_kfunc static void bbr_skb_marked_lost(struct sock *sk, ++ const struct sk_buff *skb) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ struct tcp_skb_cb *scb = TCP_SKB_CB(skb); ++ struct rate_sample rs = {}; ++ ++ bbr_note_loss(sk); ++ ++ if (!bbr->bw_probe_samples) ++ return; /* not an skb sent while probing for bandwidth */ ++ if (unlikely(!scb->tx.delivered_mstamp)) ++ return; /* skb was SACKed, reneged, marked lost; ignore it */ ++ /* We are probing for bandwidth. Construct a rate sample that ++ * estimates what happened in the flight leading up to this lost skb, ++ * then see if the loss rate went too high, and if so at which packet. ++ */ ++ rs.tx_in_flight = scb->tx.in_flight; ++ rs.lost = tp->lost - scb->tx.lost; ++ rs.is_app_limited = scb->tx.is_app_limited; ++ if (bbr_is_inflight_too_high(sk, &rs)) { ++ rs.tx_in_flight = bbr_inflight_hi_from_lost_skb(sk, &rs, skb); ++ bbr_handle_inflight_too_high(sk, &rs); ++ } ++} ++ ++static void bbr_run_loss_probe_recovery(struct sock *sk) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ struct bbr *bbr = inet_csk_ca(sk); ++ struct rate_sample rs = {0}; ++ ++ bbr_note_loss(sk); ++ ++ if (!bbr->bw_probe_samples) ++ return; /* not sent while probing for bandwidth */ ++ /* We are probing for bandwidth. Construct a rate sample that ++ * estimates what happened in the flight leading up to this ++ * loss, then see if the loss rate went too high. ++ */ ++ rs.lost = 1; /* TLP probe repaired loss of a single segment */ ++ rs.tx_in_flight = bbr->inflight_latest + rs.lost; ++ rs.is_app_limited = tp->tlp_orig_data_app_limited; ++ if (bbr_is_inflight_too_high(sk, &rs)) ++ bbr_handle_inflight_too_high(sk, &rs); ++} ++ ++/* Revert short-term model if current loss recovery event was spurious. */ ++__bpf_kfunc static u32 bbr_undo_cwnd(struct sock *sk) ++{ ++ struct bbr *bbr = inet_csk_ca(sk); ++ ++ bbr_reset_full_bw(sk); /* spurious slow-down; reset full bw detector */ ++ bbr->loss_in_round = 0; ++ ++ /* Revert to cwnd and other state saved before loss episode. */ ++ bbr->bw_lo = max(bbr->bw_lo, bbr->undo_bw_lo); ++ bbr->inflight_lo = max(bbr->inflight_lo, bbr->undo_inflight_lo); ++ bbr->inflight_hi = max(bbr->inflight_hi, bbr->undo_inflight_hi); ++ bbr->try_fast_path = 0; /* take slow path to set proper cwnd, pacing */ ++ return bbr->prior_cwnd; ++} ++ ++/* Entering loss recovery, so save state for when we undo recovery. */ + __bpf_kfunc static u32 bbr_ssthresh(struct sock *sk) + { ++ struct bbr *bbr = inet_csk_ca(sk); ++ + bbr_save_cwnd(sk); ++ /* For undo, save state that adapts based on loss signal. */ ++ bbr->undo_bw_lo = bbr->bw_lo; ++ bbr->undo_inflight_lo = bbr->inflight_lo; ++ bbr->undo_inflight_hi = bbr->inflight_hi; + return tcp_sk(sk)->snd_ssthresh; + } + ++static enum tcp_bbr_phase bbr_get_phase(struct bbr *bbr) ++{ ++ switch (bbr->mode) { ++ case BBR_STARTUP: ++ return BBR_PHASE_STARTUP; ++ case BBR_DRAIN: ++ return BBR_PHASE_DRAIN; ++ case BBR_PROBE_BW: ++ break; ++ case BBR_PROBE_RTT: ++ return BBR_PHASE_PROBE_RTT; ++ default: ++ return BBR_PHASE_INVALID; ++ } ++ switch (bbr->cycle_idx) { ++ case BBR_BW_PROBE_UP: ++ return BBR_PHASE_PROBE_BW_UP; ++ case BBR_BW_PROBE_DOWN: ++ return BBR_PHASE_PROBE_BW_DOWN; ++ case BBR_BW_PROBE_CRUISE: ++ return BBR_PHASE_PROBE_BW_CRUISE; ++ case BBR_BW_PROBE_REFILL: ++ return BBR_PHASE_PROBE_BW_REFILL; ++ default: ++ return BBR_PHASE_INVALID; ++ } ++} ++ + static size_t bbr_get_info(struct sock *sk, u32 ext, int *attr, +- union tcp_cc_info *info) ++ union tcp_cc_info *info) + { + if (ext & (1 << (INET_DIAG_BBRINFO - 1)) || + ext & (1 << (INET_DIAG_VEGASINFO - 1))) { +- struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); +- u64 bw = bbr_bw(sk); +- +- bw = bw * tp->mss_cache * USEC_PER_SEC >> BW_SCALE; +- memset(&info->bbr, 0, sizeof(info->bbr)); +- info->bbr.bbr_bw_lo = (u32)bw; +- info->bbr.bbr_bw_hi = (u32)(bw >> 32); +- info->bbr.bbr_min_rtt = bbr->min_rtt_us; +- info->bbr.bbr_pacing_gain = bbr->pacing_gain; +- info->bbr.bbr_cwnd_gain = bbr->cwnd_gain; ++ u64 bw = bbr_bw_bytes_per_sec(sk, bbr_bw(sk)); ++ u64 bw_hi = bbr_bw_bytes_per_sec(sk, bbr_max_bw(sk)); ++ u64 bw_lo = bbr->bw_lo == ~0U ? ++ ~0ULL : bbr_bw_bytes_per_sec(sk, bbr->bw_lo); ++ struct tcp_bbr_info *bbr_info = &info->bbr; ++ ++ memset(bbr_info, 0, sizeof(*bbr_info)); ++ bbr_info->bbr_bw_lo = (u32)bw; ++ bbr_info->bbr_bw_hi = (u32)(bw >> 32); ++ bbr_info->bbr_min_rtt = bbr->min_rtt_us; ++ bbr_info->bbr_pacing_gain = bbr->pacing_gain; ++ bbr_info->bbr_cwnd_gain = bbr->cwnd_gain; ++ bbr_info->bbr_bw_hi_lsb = (u32)bw_hi; ++ bbr_info->bbr_bw_hi_msb = (u32)(bw_hi >> 32); ++ bbr_info->bbr_bw_lo_lsb = (u32)bw_lo; ++ bbr_info->bbr_bw_lo_msb = (u32)(bw_lo >> 32); ++ bbr_info->bbr_mode = bbr->mode; ++ bbr_info->bbr_phase = (__u8)bbr_get_phase(bbr); ++ bbr_info->bbr_version = (__u8)BBR_VERSION; ++ bbr_info->bbr_inflight_lo = bbr->inflight_lo; ++ bbr_info->bbr_inflight_hi = bbr->inflight_hi; ++ bbr_info->bbr_extra_acked = bbr_extra_acked(sk); + *attr = INET_DIAG_BBRINFO; +- return sizeof(info->bbr); ++ return sizeof(*bbr_info); + } + return 0; + } + + __bpf_kfunc static void bbr_set_state(struct sock *sk, u8 new_state) + { ++ struct tcp_sock *tp = tcp_sk(sk); + struct bbr *bbr = inet_csk_ca(sk); + + if (new_state == TCP_CA_Loss) { +- struct rate_sample rs = { .losses = 1 }; + + bbr->prev_ca_state = TCP_CA_Loss; +- bbr->full_bw = 0; +- bbr->round_start = 1; /* treat RTO like end of a round */ +- bbr_lt_bw_sampling(sk, &rs); ++ tcp_plb_update_state_upon_rto(sk, &bbr->plb); ++ /* The tcp_write_timeout() call to sk_rethink_txhash() likely ++ * repathed this flow, so re-learn the min network RTT on the ++ * new path: ++ */ ++ bbr_reset_full_bw(sk); ++ if (!bbr_is_probing_bandwidth(sk) && bbr->inflight_lo == ~0U) { ++ /* bbr_adapt_lower_bounds() needs cwnd before ++ * we suffered an RTO, to update inflight_lo: ++ */ ++ bbr->inflight_lo = ++ max(tcp_snd_cwnd(tp), bbr->prior_cwnd); ++ } ++ } else if (bbr->prev_ca_state == TCP_CA_Loss && ++ new_state != TCP_CA_Loss) { ++ bbr_exit_loss_recovery(sk); + } + } + ++ + static struct tcp_congestion_ops tcp_bbr_cong_ops __read_mostly = { +- .flags = TCP_CONG_NON_RESTRICTED, ++ .flags = TCP_CONG_NON_RESTRICTED | TCP_CONG_WANTS_CE_EVENTS, + .name = "bbr", + .owner = THIS_MODULE, + .init = bbr_init, + .cong_control = bbr_main, + .sndbuf_expand = bbr_sndbuf_expand, ++ .skb_marked_lost = bbr_skb_marked_lost, + .undo_cwnd = bbr_undo_cwnd, + .cwnd_event = bbr_cwnd_event, + .ssthresh = bbr_ssthresh, +- .min_tso_segs = bbr_min_tso_segs, ++ .tso_segs = bbr_tso_segs, + .get_info = bbr_get_info, + .set_state = bbr_set_state, + }; +@@ -1159,10 +2359,11 @@ BTF_KFUNCS_START(tcp_bbr_check_kfunc_ids) + BTF_ID_FLAGS(func, bbr_init) + BTF_ID_FLAGS(func, bbr_main) + BTF_ID_FLAGS(func, bbr_sndbuf_expand) ++BTF_ID_FLAGS(func, bbr_skb_marked_lost) + BTF_ID_FLAGS(func, bbr_undo_cwnd) + BTF_ID_FLAGS(func, bbr_cwnd_event) + BTF_ID_FLAGS(func, bbr_ssthresh) +-BTF_ID_FLAGS(func, bbr_min_tso_segs) ++BTF_ID_FLAGS(func, bbr_tso_segs) + BTF_ID_FLAGS(func, bbr_set_state) + BTF_KFUNCS_END(tcp_bbr_check_kfunc_ids) + +@@ -1195,5 +2396,12 @@ MODULE_AUTHOR("Van Jacobson "); + MODULE_AUTHOR("Neal Cardwell "); + MODULE_AUTHOR("Yuchung Cheng "); + MODULE_AUTHOR("Soheil Hassas Yeganeh "); ++MODULE_AUTHOR("Priyaranjan Jha "); ++MODULE_AUTHOR("Yousuk Seung "); ++MODULE_AUTHOR("Kevin Yang "); ++MODULE_AUTHOR("Arjun Roy "); ++MODULE_AUTHOR("David Morley "); ++ + MODULE_LICENSE("Dual BSD/GPL"); + MODULE_DESCRIPTION("TCP BBR (Bottleneck Bandwidth and RTT)"); ++MODULE_VERSION(__stringify(BBR_VERSION)); +diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c +index 0306d257fa64..28f581c0dab7 100644 +--- a/net/ipv4/tcp_cong.c ++++ b/net/ipv4/tcp_cong.c +@@ -237,6 +237,7 @@ void tcp_init_congestion_control(struct sock *sk) + struct inet_connection_sock *icsk = inet_csk(sk); + + tcp_sk(sk)->prior_ssthresh = 0; ++ tcp_sk(sk)->fast_ack_mode = 0; + if (icsk->icsk_ca_ops->init) + icsk->icsk_ca_ops->init(sk); + if (tcp_ca_needs_ecn(sk)) +diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c +index cc05ec1faac8..7e995b9ea034 100644 +--- a/net/ipv4/tcp_input.c ++++ b/net/ipv4/tcp_input.c +@@ -370,7 +370,7 @@ static void __tcp_ecn_check_ce(struct sock *sk, const struct sk_buff *skb) + tcp_enter_quickack_mode(sk, 2); + break; + case INET_ECN_CE: +- if (tcp_ca_needs_ecn(sk)) ++ if (tcp_ca_wants_ce_events(sk)) + tcp_ca_event(sk, CA_EVENT_ECN_IS_CE); + + if (!(tp->ecn_flags & TCP_ECN_DEMAND_CWR)) { +@@ -381,7 +381,7 @@ static void __tcp_ecn_check_ce(struct sock *sk, const struct sk_buff *skb) + tp->ecn_flags |= TCP_ECN_SEEN; + break; + default: +- if (tcp_ca_needs_ecn(sk)) ++ if (tcp_ca_wants_ce_events(sk)) + tcp_ca_event(sk, CA_EVENT_ECN_NO_CE); + tp->ecn_flags |= TCP_ECN_SEEN; + break; +@@ -1120,7 +1120,12 @@ static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb) + */ + static void tcp_notify_skb_loss_event(struct tcp_sock *tp, const struct sk_buff *skb) + { ++ struct sock *sk = (struct sock *)tp; ++ const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops; ++ + tp->lost += tcp_skb_pcount(skb); ++ if (ca_ops->skb_marked_lost) ++ ca_ops->skb_marked_lost(sk, skb); + } + + void tcp_mark_skb_lost(struct sock *sk, struct sk_buff *skb) +@@ -1501,6 +1506,17 @@ static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *prev, + WARN_ON_ONCE(tcp_skb_pcount(skb) < pcount); + tcp_skb_pcount_add(skb, -pcount); + ++ /* Adjust tx.in_flight as pcount is shifted from skb to prev. */ ++ if (WARN_ONCE(TCP_SKB_CB(skb)->tx.in_flight < pcount, ++ "prev in_flight: %u skb in_flight: %u pcount: %u", ++ TCP_SKB_CB(prev)->tx.in_flight, ++ TCP_SKB_CB(skb)->tx.in_flight, ++ pcount)) ++ TCP_SKB_CB(skb)->tx.in_flight = 0; ++ else ++ TCP_SKB_CB(skb)->tx.in_flight -= pcount; ++ TCP_SKB_CB(prev)->tx.in_flight += pcount; ++ + /* When we're adding to gso_segs == 1, gso_size will be zero, + * in theory this shouldn't be necessary but as long as DSACK + * code can come after this skb later on it's better to keep +@@ -3799,7 +3815,8 @@ static void tcp_replace_ts_recent(struct tcp_sock *tp, u32 seq) + /* This routine deals with acks during a TLP episode and ends an episode by + * resetting tlp_high_seq. Ref: TLP algorithm in draft-ietf-tcpm-rack + */ +-static void tcp_process_tlp_ack(struct sock *sk, u32 ack, int flag) ++static void tcp_process_tlp_ack(struct sock *sk, u32 ack, int flag, ++ struct rate_sample *rs) + { + struct tcp_sock *tp = tcp_sk(sk); + +@@ -3816,6 +3833,7 @@ static void tcp_process_tlp_ack(struct sock *sk, u32 ack, int flag) + /* ACK advances: there was a loss, so reduce cwnd. Reset + * tlp_high_seq in tcp_init_cwnd_reduction() + */ ++ tcp_ca_event(sk, CA_EVENT_TLP_RECOVERY); + tcp_init_cwnd_reduction(sk); + tcp_set_ca_state(sk, TCP_CA_CWR); + tcp_end_cwnd_reduction(sk); +@@ -3826,6 +3844,11 @@ static void tcp_process_tlp_ack(struct sock *sk, u32 ack, int flag) + FLAG_NOT_DUP | FLAG_DATA_SACKED))) { + /* Pure dupack: original and TLP probe arrived; no loss */ + tp->tlp_high_seq = 0; ++ } else { ++ /* This ACK matches a TLP retransmit. We cannot yet tell if ++ * this ACK is for the original or the TLP retransmit. ++ */ ++ rs->is_acking_tlp_retrans_seq = 1; + } + } + +@@ -3934,6 +3957,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) + + prior_fack = tcp_is_sack(tp) ? tcp_highest_sack_seq(tp) : tp->snd_una; + rs.prior_in_flight = tcp_packets_in_flight(tp); ++ tcp_rate_check_app_limited(sk); + + /* ts_recent update must be made after we are sure that the packet + * is in window. +@@ -4008,7 +4032,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) + tcp_rack_update_reo_wnd(sk, &rs); + + if (tp->tlp_high_seq) +- tcp_process_tlp_ack(sk, ack, flag); ++ tcp_process_tlp_ack(sk, ack, flag, &rs); + + if (tcp_ack_is_dubious(sk, flag)) { + if (!(flag & (FLAG_SND_UNA_ADVANCED | +@@ -4032,6 +4056,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) + delivered = tcp_newly_delivered(sk, delivered, flag); + lost = tp->lost - lost; /* freshly marked lost */ + rs.is_ack_delayed = !!(flag & FLAG_ACK_MAYBE_DELAYED); ++ rs.is_ece = !!(flag & FLAG_ECE); + tcp_rate_gen(sk, delivered, lost, is_sack_reneg, sack_state.rate); + tcp_cong_control(sk, ack, delivered, flag, sack_state.rate); + tcp_xmit_recovery(sk, rexmit); +@@ -4051,7 +4076,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) + tcp_ack_probe(sk); + + if (tp->tlp_high_seq) +- tcp_process_tlp_ack(sk, ack, flag); ++ tcp_process_tlp_ack(sk, ack, flag, &rs); + return 1; + + old_ack: +@@ -5725,13 +5750,14 @@ static void __tcp_ack_snd_check(struct sock *sk, int ofo_possible) + + /* More than one full frame received... */ + if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss && ++ (tp->fast_ack_mode == 1 || + /* ... and right edge of window advances far enough. + * (tcp_recvmsg() will send ACK otherwise). + * If application uses SO_RCVLOWAT, we want send ack now if + * we have not received enough bytes to satisfy the condition. + */ +- (tp->rcv_nxt - tp->copied_seq < sk->sk_rcvlowat || +- __tcp_select_window(sk) >= tp->rcv_wnd)) || ++ (tp->rcv_nxt - tp->copied_seq < sk->sk_rcvlowat || ++ __tcp_select_window(sk) >= tp->rcv_wnd))) || + /* We ACK each frame or... */ + tcp_in_quickack_mode(sk) || + /* Protocol state mandates a one-time immediate ACK */ +diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c +index bb1fe1ba867a..050a80769de6 100644 +--- a/net/ipv4/tcp_minisocks.c ++++ b/net/ipv4/tcp_minisocks.c +@@ -462,6 +462,8 @@ void tcp_ca_openreq_child(struct sock *sk, const struct dst_entry *dst) + u32 ca_key = dst_metric(dst, RTAX_CC_ALGO); + bool ca_got_dst = false; + ++ tcp_set_ecn_low_from_dst(sk, dst); ++ + if (ca_key != TCP_CA_UNSPEC) { + const struct tcp_congestion_ops *ca; + +diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c +index 4fd746bd4d54..273fed558494 100644 +--- a/net/ipv4/tcp_output.c ++++ b/net/ipv4/tcp_output.c +@@ -336,10 +336,9 @@ static void tcp_ecn_send_syn(struct sock *sk, struct sk_buff *skb) + bool bpf_needs_ecn = tcp_bpf_ca_needs_ecn(sk); + bool use_ecn = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_ecn) == 1 || + tcp_ca_needs_ecn(sk) || bpf_needs_ecn; ++ const struct dst_entry *dst = __sk_dst_get(sk); + + if (!use_ecn) { +- const struct dst_entry *dst = __sk_dst_get(sk); +- + if (dst && dst_feature(dst, RTAX_FEATURE_ECN)) + use_ecn = true; + } +@@ -351,6 +350,9 @@ static void tcp_ecn_send_syn(struct sock *sk, struct sk_buff *skb) + tp->ecn_flags = TCP_ECN_OK; + if (tcp_ca_needs_ecn(sk) || bpf_needs_ecn) + INET_ECN_xmit(sk); ++ ++ if (dst) ++ tcp_set_ecn_low_from_dst(sk, dst); + } + } + +@@ -388,7 +390,8 @@ static void tcp_ecn_send(struct sock *sk, struct sk_buff *skb, + th->cwr = 1; + skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; + } +- } else if (!tcp_ca_needs_ecn(sk)) { ++ } else if (!(tp->ecn_flags & TCP_ECN_ECT_PERMANENT) && ++ !tcp_ca_needs_ecn(sk)) { + /* ACK or retransmitted segment: clear ECT|CE */ + INET_ECN_dontxmit(sk); + } +@@ -1601,7 +1604,7 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue, + { + struct tcp_sock *tp = tcp_sk(sk); + struct sk_buff *buff; +- int old_factor; ++ int old_factor, inflight_prev; + long limit; + int nlen; + u8 flags; +@@ -1676,6 +1679,30 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue, + + if (diff) + tcp_adjust_pcount(sk, skb, diff); ++ ++ inflight_prev = TCP_SKB_CB(skb)->tx.in_flight - old_factor; ++ if (inflight_prev < 0) { ++ WARN_ONCE(tcp_skb_tx_in_flight_is_suspicious( ++ old_factor, ++ TCP_SKB_CB(skb)->sacked, ++ TCP_SKB_CB(skb)->tx.in_flight), ++ "inconsistent: tx.in_flight: %u " ++ "old_factor: %d mss: %u sacked: %u " ++ "1st pcount: %d 2nd pcount: %d " ++ "1st len: %u 2nd len: %u ", ++ TCP_SKB_CB(skb)->tx.in_flight, old_factor, ++ mss_now, TCP_SKB_CB(skb)->sacked, ++ tcp_skb_pcount(skb), tcp_skb_pcount(buff), ++ skb->len, buff->len); ++ inflight_prev = 0; ++ } ++ /* Set 1st tx.in_flight as if 1st were sent by itself: */ ++ TCP_SKB_CB(skb)->tx.in_flight = inflight_prev + ++ tcp_skb_pcount(skb); ++ /* Set 2nd tx.in_flight with new 1st and 2nd pcounts: */ ++ TCP_SKB_CB(buff)->tx.in_flight = inflight_prev + ++ tcp_skb_pcount(skb) + ++ tcp_skb_pcount(buff); + } + + /* Link BUFF into the send queue. */ +@@ -2033,13 +2060,12 @@ static u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now, + static u32 tcp_tso_segs(struct sock *sk, unsigned int mss_now) + { + const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops; +- u32 min_tso, tso_segs; +- +- min_tso = ca_ops->min_tso_segs ? +- ca_ops->min_tso_segs(sk) : +- READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_min_tso_segs); ++ u32 tso_segs; + +- tso_segs = tcp_tso_autosize(sk, mss_now, min_tso); ++ tso_segs = ca_ops->tso_segs ? ++ ca_ops->tso_segs(sk, mss_now) : ++ tcp_tso_autosize(sk, mss_now, ++ sock_net(sk)->ipv4.sysctl_tcp_min_tso_segs); + return min_t(u32, tso_segs, sk->sk_gso_max_segs); + } + +@@ -2768,6 +2794,7 @@ static bool tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle, + skb_set_delivery_time(skb, tp->tcp_wstamp_ns, SKB_CLOCK_MONOTONIC); + list_move_tail(&skb->tcp_tsorted_anchor, &tp->tsorted_sent_queue); + tcp_init_tso_segs(skb, mss_now); ++ tcp_set_tx_in_flight(sk, skb); + goto repair; /* Skip network transmission */ + } + +@@ -2982,6 +3009,7 @@ void tcp_send_loss_probe(struct sock *sk) + if (WARN_ON(!skb || !tcp_skb_pcount(skb))) + goto rearm_timer; + ++ tp->tlp_orig_data_app_limited = TCP_SKB_CB(skb)->tx.is_app_limited; + if (__tcp_retransmit_skb(sk, skb, 1)) + goto rearm_timer; + +diff --git a/net/ipv4/tcp_rate.c b/net/ipv4/tcp_rate.c +index a8f6d9d06f2e..8737f2134648 100644 +--- a/net/ipv4/tcp_rate.c ++++ b/net/ipv4/tcp_rate.c +@@ -34,6 +34,24 @@ + * ready to send in the write queue. + */ + ++void tcp_set_tx_in_flight(struct sock *sk, struct sk_buff *skb) ++{ ++ struct tcp_sock *tp = tcp_sk(sk); ++ u32 in_flight; ++ ++ /* Check, sanitize, and record packets in flight after skb was sent. */ ++ in_flight = tcp_packets_in_flight(tp) + tcp_skb_pcount(skb); ++ if (WARN_ONCE(in_flight > TCPCB_IN_FLIGHT_MAX, ++ "insane in_flight %u cc %s mss %u " ++ "cwnd %u pif %u %u %u %u\n", ++ in_flight, inet_csk(sk)->icsk_ca_ops->name, ++ tp->mss_cache, tp->snd_cwnd, ++ tp->packets_out, tp->retrans_out, ++ tp->sacked_out, tp->lost_out)) ++ in_flight = TCPCB_IN_FLIGHT_MAX; ++ TCP_SKB_CB(skb)->tx.in_flight = in_flight; ++} ++ + /* Snapshot the current delivery information in the skb, to generate + * a rate sample later when the skb is (s)acked in tcp_rate_skb_delivered(). + */ +@@ -66,7 +84,9 @@ void tcp_rate_skb_sent(struct sock *sk, struct sk_buff *skb) + TCP_SKB_CB(skb)->tx.delivered_mstamp = tp->delivered_mstamp; + TCP_SKB_CB(skb)->tx.delivered = tp->delivered; + TCP_SKB_CB(skb)->tx.delivered_ce = tp->delivered_ce; ++ TCP_SKB_CB(skb)->tx.lost = tp->lost; + TCP_SKB_CB(skb)->tx.is_app_limited = tp->app_limited ? 1 : 0; ++ tcp_set_tx_in_flight(sk, skb); + } + + /* When an skb is sacked or acked, we fill in the rate sample with the (prior) +@@ -91,18 +111,21 @@ void tcp_rate_skb_delivered(struct sock *sk, struct sk_buff *skb, + if (!rs->prior_delivered || + tcp_skb_sent_after(tx_tstamp, tp->first_tx_mstamp, + scb->end_seq, rs->last_end_seq)) { ++ rs->prior_lost = scb->tx.lost; + rs->prior_delivered_ce = scb->tx.delivered_ce; + rs->prior_delivered = scb->tx.delivered; + rs->prior_mstamp = scb->tx.delivered_mstamp; + rs->is_app_limited = scb->tx.is_app_limited; + rs->is_retrans = scb->sacked & TCPCB_RETRANS; ++ rs->tx_in_flight = scb->tx.in_flight; + rs->last_end_seq = scb->end_seq; + + /* Record send time of most recently ACKed packet: */ + tp->first_tx_mstamp = tx_tstamp; + /* Find the duration of the "send phase" of this window: */ +- rs->interval_us = tcp_stamp_us_delta(tp->first_tx_mstamp, +- scb->tx.first_tx_mstamp); ++ rs->interval_us = tcp_stamp32_us_delta( ++ tp->first_tx_mstamp, ++ scb->tx.first_tx_mstamp); + + } + /* Mark off the skb delivered once it's sacked to avoid being +@@ -144,6 +167,7 @@ void tcp_rate_gen(struct sock *sk, u32 delivered, u32 lost, + return; + } + rs->delivered = tp->delivered - rs->prior_delivered; ++ rs->lost = tp->lost - rs->prior_lost; + + rs->delivered_ce = tp->delivered_ce - rs->prior_delivered_ce; + /* delivered_ce occupies less than 32 bits in the skb control block */ +@@ -155,7 +179,7 @@ void tcp_rate_gen(struct sock *sk, u32 delivered, u32 lost, + * longer phase. + */ + snd_us = rs->interval_us; /* send phase */ +- ack_us = tcp_stamp_us_delta(tp->tcp_mstamp, ++ ack_us = tcp_stamp32_us_delta(tp->tcp_mstamp, + rs->prior_mstamp); /* ack phase */ + rs->interval_us = max(snd_us, ack_us); + +diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c +index 79064580c8c0..697270ce1ea6 100644 +--- a/net/ipv4/tcp_timer.c ++++ b/net/ipv4/tcp_timer.c +@@ -690,6 +690,7 @@ void tcp_write_timer_handler(struct sock *sk) + return; + } + ++ tcp_rate_check_app_limited(sk); + tcp_mstamp_refresh(tcp_sk(sk)); + event = icsk->icsk_pending; + +-- +2.47.0.rc0 + diff --git a/patches/sys-kernel/gentoo-sources/0003-glitched-base.patch b/patches/sys-kernel/gentoo-sources/0003-glitched-base.patch new file mode 100644 index 0000000..beb4aae --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0003-glitched-base.patch @@ -0,0 +1,822 @@ +From f7f49141a5dbe9c99d78196b58c44307fb2e6be3 Mon Sep 17 00:00:00 2001 +From: Tk-Glitch +Date: Wed, 4 Jul 2018 04:30:08 +0200 +Subject: [PATCH 01/17] glitched + +--- + init/Makefile | 2 +- + 1 file changed, 1 insertions(+), 1 deletions(-) + +diff --git a/init/Makefile b/init/Makefile +index baf3ab8d9d49..854e32e6aec7 100755 +--- a/init/Makefile ++++ b/init/Makefile +@@ -19,7 +19,7 @@ else + + # Maximum length of UTS_VERSION is 64 chars + filechk_uts_version = \ +- utsver=$$(echo '$(pound)'"$(build-version)" $(smp-flag-y) $(preempt-flag-y) "$(build-timestamp)" | cut -b -64); \ ++ utsver=$$(echo '$(pound)'"$(build-version)" $(smp-flag-y) $(preempt-flag-y) "siina" "$(build-timestamp)" | cut -b -64); \ + echo '$(pound)'define UTS_VERSION \""$${utsver}"\" + + # +-- +2.28.0 + + +From c304f43d14e98d4bf1215fc10bc5012f554bdd8a Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 29 Jan 2018 16:59:22 +0000 +Subject: [PATCH 02/17] dcache: cache_pressure = 50 decreases the rate at which + VFS caches are reclaimed + +Signed-off-by: Alexandre Frade +--- + fs/dcache.c | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/fs/dcache.c b/fs/dcache.c +index 361ea7ab30ea..0c5cf69b241a 100644 +--- a/fs/dcache.c ++++ b/fs/dcache.c +@@ -71,7 +71,7 @@ + * If no ancestor relationship: + * arbitrary, since it's serialized on rename_lock + */ +-int sysctl_vfs_cache_pressure __read_mostly = 100; ++int sysctl_vfs_cache_pressure __read_mostly = 50; + EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure); + + __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock); +-- +2.28.0 + + +diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c +index f788cd61df21..2bfbb4213707 100644 +--- a/kernel/sched/rt.c ++++ b/kernel/sched/rt.c +@@ -15,9 +15,9 @@ __read_mostly int scheduler_running; + + /* + * part of the period that we allow rt tasks to run in us. +- * default: 0.95s ++ * XanMod default: 0.98s + */ +-int sysctl_sched_rt_runtime = 950000; ++int sysctl_sched_rt_runtime = 980000; + + #ifdef CONFIG_SYSCTL + static int sysctl_sched_rr_timeslice = (MSEC_PER_SEC / HZ) * RR_TIMESLICE; +-- +2.28.0 + + +From acc49f33a10f61dc66c423888cbb883ba46710e4 Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 29 Jan 2018 17:41:29 +0000 +Subject: [PATCH 04/17] scripts: disable the localversion "+" tag of a git repo + +Signed-off-by: Alexandre Frade +--- + scripts/setlocalversion | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/scripts/setlocalversion b/scripts/setlocalversion +index 20f2efd57b11..0552d8b9f582 100755 +--- a/scripts/setlocalversion ++++ b/scripts/setlocalversion +@@ -54,7 +54,7 @@ scm_version() + # If only the short version is requested, don't bother + # running further git commands + if $short; then +- echo "+" ++ #echo "+" + return + fi + # If we are past the tagged commit, we pretty print it. +-- +2.28.0 + + +From 360c6833e07cc9fdef5746f6bc45bdbc7212288d Mon Sep 17 00:00:00 2001 +From: "Jan Alexander Steffens (heftig)" +Date: Fri, 26 Oct 2018 11:22:33 +0100 +Subject: [PATCH 06/17] infiniband: Fix __read_overflow2 error with -O3 + inlining + +--- + drivers/infiniband/core/addr.c | 1 + + 1 file changed, 1 insertion(+) + +diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c +index 3a98439bba83..6efc4f907f58 100644 +--- a/drivers/infiniband/core/addr.c ++++ b/drivers/infiniband/core/addr.c +@@ -820,6 +820,7 @@ int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, + union { + struct sockaddr_in _sockaddr_in; + struct sockaddr_in6 _sockaddr_in6; ++ struct sockaddr_ib _sockaddr_ib; + } sgid_addr, dgid_addr; + int ret; + +-- +2.28.0 + + +From f85ed068b4d0e6c31edce8574a95757a60e58b87 Mon Sep 17 00:00:00 2001 +From: Etienne Juvigny +Date: Mon, 3 Sep 2018 17:36:25 +0200 +Subject: [PATCH 07/17] Add Zenify option + +--- + init/Kconfig | 32 ++++++++++++++++++++++++++++++++ + 1 file changed, 32 insertions(+) + +diff --git a/init/Kconfig b/init/Kconfig +index 3ae8678e1145..da708eed0f1e 100644 +--- a/init/Kconfig ++++ b/init/Kconfig +@@ -92,6 +92,38 @@ config THREAD_INFO_IN_TASK + + menu "General setup" + ++config ZENIFY ++ bool "A selection of patches from Zen/Liquorix kernel and additional tweaks for a better gaming experience" ++ default y ++ help ++ Tunes the kernel for responsiveness at the cost of throughput and power usage. ++ ++ --- Virtual Memory Subsystem --------------------------- ++ ++ Mem dirty before bg writeback..: 10 % -> 20 % ++ Mem dirty before sync writeback: 20 % -> 50 % ++ ++ --- Block Layer ---------------------------------------- ++ ++ Queue depth...............: 128 -> 512 ++ Default MQ scheduler......: mq-deadline -> bfq ++ ++ --- CFS CPU Scheduler ---------------------------------- ++ ++ Scheduling latency.............: 6 -> 3 ms ++ Minimal granularity............: 0.75 -> 0.3 ms ++ Wakeup granularity.............: 1 -> 0.5 ms ++ CPU migration cost.............: 0.5 -> 0.25 ms ++ Bandwidth slice size...........: 5 -> 3 ms ++ Ondemand fine upscaling limit..: 95 % -> 85 % ++ ++ --- MuQSS CPU Scheduler -------------------------------- ++ ++ Scheduling interval............: 6 -> 3 ms ++ ISO task max realtime use......: 70 % -> 25 % ++ Ondemand coarse upscaling limit: 80 % -> 45 % ++ Ondemand fine upscaling limit..: 95 % -> 45 % ++ + config BROKEN + bool + +-- +2.28.0 + + +From e92e67143385cf285851e12aa8b7f083dd38dd24 Mon Sep 17 00:00:00 2001 +From: Steven Barrett +Date: Sun, 16 Jan 2011 18:57:32 -0600 +Subject: [PATCH 08/17] ZEN: Allow TCP YeAH as default congestion control + +4.4: In my tests YeAH dramatically slowed down transfers over a WLAN, + reducing throughput from ~65Mbps (CUBIC) to ~7MBps (YeAH) over 10 + seconds (netperf TCP_STREAM) including long stalls. + + Be careful when choosing this. ~heftig +--- + net/ipv4/Kconfig | 4 ++++ + 1 file changed, 4 insertions(+) + +diff --git a/net/ipv4/Kconfig b/net/ipv4/Kconfig +index e64e59b536d3..bfb55ef7ebbe 100644 +--- a/net/ipv4/Kconfig ++++ b/net/ipv4/Kconfig +@@ -691,6 +691,9 @@ choice + config DEFAULT_VEGAS + bool "Vegas" if TCP_CONG_VEGAS=y + ++ config DEFAULT_YEAH ++ bool "YeAH" if TCP_CONG_YEAH=y ++ + config DEFAULT_VENO + bool "Veno" if TCP_CONG_VENO=y + +@@ -724,6 +727,7 @@ config DEFAULT_TCP_CONG + default "htcp" if DEFAULT_HTCP + default "hybla" if DEFAULT_HYBLA + default "vegas" if DEFAULT_VEGAS ++ default "yeah" if DEFAULT_YEAH + default "westwood" if DEFAULT_WESTWOOD + default "veno" if DEFAULT_VENO + default "reno" if DEFAULT_RENO +-- +2.28.0 + + +From 76dbe7477bfde1b5e8bf29a71b5af7ab2be9b98e Mon Sep 17 00:00:00 2001 +From: Steven Barrett +Date: Wed, 28 Nov 2018 19:01:27 -0600 +Subject: [PATCH 09/17] zen: Use [defer+madvise] as default khugepaged defrag + strategy + +For some reason, the default strategy to respond to THP fault fallbacks +is still just madvise, meaning stall if the program wants transparent +hugepages, but don't trigger a background reclaim / compaction if THP +begins to fail allocations. This creates a snowball affect where we +still use the THP code paths, but we almost always fail once a system +has been active and busy for a while. + +The option "defer" was created for interactive systems where THP can +still improve performance. If we have to fallback to a regular page due +to an allocation failure or anything else, we will trigger a background +reclaim and compaction so future THP attempts succeed and previous +attempts eventually have their smaller pages combined without stalling +running applications. + +We still want madvise to stall applications that explicitely want THP, +so defer+madvise _does_ make a ton of sense. Make it the default for +interactive systems, especially if the kernel maintainer left +transparent hugepages on "always". + +Reasoning and details in the original patch: https://lwn.net/Articles/711248/ +--- + mm/huge_memory.c | 4 ++++ + 1 file changed, 4 insertions(+) + +diff --git a/mm/huge_memory.c b/mm/huge_memory.c +index 74300e337c3c..9277f22c10a7 100644 +--- a/mm/huge_memory.c ++++ b/mm/huge_memory.c +@@ -53,7 +53,11 @@ unsigned long transparent_hugepage_flags __read_mostly = + #ifdef CONFIG_TRANSPARENT_HUGEPAGE_MADVISE + (1< +Date: Wed, 24 Oct 2018 16:58:52 -0300 +Subject: [PATCH 10/17] net/sched: allow configuring cake qdisc as default + +Signed-off-by: Alexandre Frade +--- + net/sched/Kconfig | 4 ++++ + 1 file changed, 4 insertions(+) + +diff --git a/net/sched/Kconfig b/net/sched/Kconfig +index 84badf00647e..6a922bca9f39 100644 +--- a/net/sched/Kconfig ++++ b/net/sched/Kconfig +@@ -471,6 +471,9 @@ choice + config DEFAULT_SFQ + bool "Stochastic Fair Queue" if NET_SCH_SFQ + ++ config DEFAULT_CAKE ++ bool "Common Applications Kept Enhanced" if NET_SCH_CAKE ++ + config DEFAULT_PFIFO_FAST + bool "Priority FIFO Fast" + endchoice +@@ -481,6 +484,7 @@ config DEFAULT_NET_SCH + default "fq" if DEFAULT_FQ + default "fq_codel" if DEFAULT_FQ_CODEL + default "sfq" if DEFAULT_SFQ ++ default "cake" if DEFAULT_CAKE + default "pfifo_fast" + endif + +-- +2.28.0 + + +From 90240bcd90a568878738e66c0d45bed3e38e347b Mon Sep 17 00:00:00 2001 +From: Tk-Glitch +Date: Fri, 19 Apr 2019 12:33:38 +0200 +Subject: [PATCH 12/17] Set vm.max_map_count to 262144 by default + +The value is still pretty low, and AMD64-ABI and ELF extended numbering +supports that, so we should be fine on modern x86 systems. + +This fixes crashes in some applications using more than 65535 vmas (also +affects some windows games running in wine, such as Star Citizen). +--- + include/linux/mm.h | 3 +-- + 1 file changed, 1 insertion(+), 2 deletions(-) + +diff --git a/include/linux/mm.h b/include/linux/mm.h +index bc05c3588aa3..b0cefe94920d 100644 +--- a/include/linux/mm.h ++++ b/include/linux/mm.h +@@ -190,8 +190,7 @@ static inline void __mm_zero_struct_page(struct page *page) + * not a hard limit any more. Although some userspace tools can be surprised by + * that. + */ +-#define MAPCOUNT_ELF_CORE_MARGIN (5) +-#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) ++#define DEFAULT_MAX_MAP_COUNT (262144) + + extern int sysctl_max_map_count; + +-- +2.28.0 + + +From 3a34034dba5efe91bcec491efe8c66e8087f509b Mon Sep 17 00:00:00 2001 +From: Tk-Glitch +Date: Mon, 27 Jul 2020 00:19:18 +0200 +Subject: [PATCH 13/17] mm: bump DEFAULT_MAX_MAP_COUNT + +Some games such as Detroit: Become Human tend to be very crash prone with +lower values. +--- + include/linux/mm.h | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/include/linux/mm.h b/include/linux/mm.h +index b0cefe94920d..890165099b07 100644 +--- a/include/linux/mm.h ++++ b/include/linux/mm.h +@@ -190,7 +190,7 @@ static inline void __mm_zero_struct_page(struct page *page) + * not a hard limit any more. Although some userspace tools can be surprised by + * that. + */ +-#define DEFAULT_MAX_MAP_COUNT (262144) ++#define DEFAULT_MAX_MAP_COUNT (16777216) + + extern int sysctl_max_map_count; + +-- +2.28.0 + +From 977812938da7c7226415778c340832141d9278b7 Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 25 Nov 2019 15:13:06 -0300 +Subject: [PATCH 14/17] elevator: set default scheduler to bfq for blk-mq + +Signed-off-by: Alexandre Frade +--- + block/elevator.c | 6 +++--- + 1 file changed, 3 insertions(+), 3 deletions(-) + +diff --git a/block/elevator.c b/block/elevator.c +index 4eab3d70e880..79669aa39d79 100644 +--- a/block/elevator.c ++++ b/block/elevator.c +@@ -623,19 +623,19 @@ static inline bool elv_support_iosched(struct request_queue *q) + } + + /* +- * For single queue devices, default to using mq-deadline. If we have multiple +- * queues or mq-deadline is not available, default to "none". ++ * For single queue devices, default to using bfq. If we have multiple ++ * queues or bfq is not available, default to "none". + */ + static struct elevator_type *elevator_get_default(struct request_queue *q) + { + if (q->tag_set && q->tag_set->flags & BLK_MQ_F_NO_SCHED_BY_DEFAULT) + return NULL; + + if (q->nr_hw_queues != 1 && + !blk_mq_is_shared_tags(q->tag_set->flags)) + return NULL; + +- return elevator_find_get(q, "mq-deadline"); ++ return elevator_find_get(q, "bfq"); + } + + /* +-- +2.28.0 + +From 3c229f434aca65c4ca61772bc03c3e0370817b92 Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 3 Aug 2020 17:05:04 +0000 +Subject: [PATCH 16/17] mm: set 2 megabytes for address_space-level file + read-ahead pages size + +Signed-off-by: Alexandre Frade +--- + include/linux/pagemap.h | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h +index cf2468da68e9..007dea784451 100644 +--- a/include/linux/pagemap.h ++++ b/include/linux/pagemap.h +@@ -655,7 +655,7 @@ int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask); + void delete_from_page_cache_batch(struct address_space *mapping, + struct pagevec *pvec); + +-#define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) ++#define VM_READAHEAD_PAGES (SZ_2M / PAGE_SIZE) + + void page_cache_sync_readahead(struct address_space *, struct file_ra_state *, + struct file *, pgoff_t index, unsigned long req_count); +-- +2.28.0 + + +From 716f41cf6631f3a85834dcb67b4ce99185b6387f Mon Sep 17 00:00:00 2001 +From: Steven Barrett +Date: Wed, 15 Jan 2020 20:43:56 -0600 +Subject: [PATCH 17/17] ZEN: intel-pstate: Implement "enable" parameter + +If intel-pstate is compiled into the kernel, it will preempt the loading +of acpi-cpufreq so you can take advantage of hardware p-states without +any friction. + +However, intel-pstate is not completely superior to cpufreq's ondemand +for one reason. There's no concept of an up_threshold property. + +In ondemand, up_threshold essentially reduces the maximum utilization to +compare against, allowing you to hit max frequencies and turbo boost +from a much lower core utilization. + +With intel-pstate, you have the concept of minimum and maximum +performance, but no tunable that lets you define, maximum frequency +means 50% core utilization. For just this oversight, there's reasons +you may want ondemand. + +Lets support setting "enable" in kernel boot parameters. This lets +kernel maintainers include "intel_pstate=disable" statically in the +static boot parameters, but let users of the kernel override this +selection. +--- + Documentation/admin-guide/kernel-parameters.txt | 3 +++ + drivers/cpufreq/intel_pstate.c | 2 ++ + 2 files changed, 5 insertions(+) + +diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt +index fb95fad81c79..3e92fee81e33 100644 +--- a/Documentation/admin-guide/kernel-parameters.txt ++++ b/Documentation/admin-guide/kernel-parameters.txt +@@ -1857,6 +1857,9 @@ + disable + Do not enable intel_pstate as the default + scaling driver for the supported processors ++ enable ++ Enable intel_pstate in-case "disable" was passed ++ previously in the kernel boot parameters + passive + Use intel_pstate as a scaling driver, but configure it + to work with generic cpufreq governors (instead of +diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c +index 36a469150ff9..aee891c9b78a 100644 +--- a/drivers/cpufreq/intel_pstate.c ++++ b/drivers/cpufreq/intel_pstate.c +@@ -2845,6 +2845,8 @@ static int __init intel_pstate_setup(char *str) + if (!strcmp(str, "no_hwp")) + no_hwp = 1; + ++ if (!strcmp(str, "enable")) ++ no_load = 0; + if (!strcmp(str, "force")) + force_load = 1; + if (!strcmp(str, "hwp_only")) +-- +2.28.0 + +From 379cbab18b5c75c622b93e2c5abdfac141fe9654 Mon Sep 17 00:00:00 2001 +From: Kenny Levinsen +Date: Sun, 27 Dec 2020 14:43:13 +0000 +Subject: [PATCH] ZEN: Input: evdev - use call_rcu when detaching client + +Significant time was spent on synchronize_rcu in evdev_detach_client +when applications closed evdev devices. Switching VT away from a +graphical environment commonly leads to mass input device closures, +which could lead to noticable delays on systems with many input devices. + +Replace synchronize_rcu with call_rcu, deferring reclaim of the evdev +client struct till after the RCU grace period instead of blocking the +calling application. + +While this does not solve all slow evdev fd closures, it takes care of a +good portion of them, including this simple test: + + #include + #include + + int main(int argc, char *argv[]) + { + int idx, fd; + const char *path = "/dev/input/event0"; + for (idx = 0; idx < 1000; idx++) { + if ((fd = open(path, O_RDWR)) == -1) { + return -1; + } + close(fd); + } + return 0; + } + +Time to completion of above test when run locally: + + Before: 0m27.111s + After: 0m0.018s + +Signed-off-by: Kenny Levinsen +--- + drivers/input/evdev.c | 19 +++++++++++-------- + 1 file changed, 11 insertions(+), 8 deletions(-) + +diff --git a/drivers/input/evdev.c b/drivers/input/evdev.c +index 95f90699d2b17b..2b10fe29d2c8d9 100644 +--- a/drivers/input/evdev.c ++++ b/drivers/input/evdev.c +@@ -46,6 +46,7 @@ struct evdev_client { + struct fasync_struct *fasync; + struct evdev *evdev; + struct list_head node; ++ struct rcu_head rcu; + enum input_clock_type clk_type; + bool revoked; + unsigned long *evmasks[EV_CNT]; +@@ -377,13 +378,22 @@ static void evdev_attach_client(struct evdev *evdev, + spin_unlock(&evdev->client_lock); + } + ++static void evdev_reclaim_client(struct rcu_head *rp) ++{ ++ struct evdev_client *client = container_of(rp, struct evdev_client, rcu); ++ unsigned int i; ++ for (i = 0; i < EV_CNT; ++i) ++ bitmap_free(client->evmasks[i]); ++ kvfree(client); ++} ++ + static void evdev_detach_client(struct evdev *evdev, + struct evdev_client *client) + { + spin_lock(&evdev->client_lock); + list_del_rcu(&client->node); + spin_unlock(&evdev->client_lock); +- synchronize_rcu(); ++ call_rcu(&client->rcu, evdev_reclaim_client); + } + + static int evdev_open_device(struct evdev *evdev) +@@ -436,7 +446,6 @@ static int evdev_release(struct inode *inode, struct file *file) + { + struct evdev_client *client = file->private_data; + struct evdev *evdev = client->evdev; +- unsigned int i; + + mutex_lock(&evdev->mutex); + +@@ -448,11 +457,6 @@ static int evdev_release(struct inode *inode, struct file *file) + + evdev_detach_client(evdev, client); + +- for (i = 0; i < EV_CNT; ++i) +- bitmap_free(client->evmasks[i]); +- +- kvfree(client); +- + evdev_close_device(evdev); + + return 0; +@@ -495,7 +499,6 @@ static int evdev_open(struct inode *inode, struct file *file) + + err_free_client: + evdev_detach_client(evdev, client); +- kvfree(client); + return error; + } + + +From 2aafb56f20e4b63d8c4af172fe9d017c64bc4129 Mon Sep 17 00:00:00 2001 +From: Sultan Alsawaf +Date: Wed, 20 Oct 2021 20:50:11 -0700 +Subject: [PATCH] ZEN: mm: Lower the non-hugetlbpage pageblock size to reduce + scheduling delays + +The page allocator processes free pages in groups of pageblocks, where +the size of a pageblock is typically quite large (1024 pages without +hugetlbpage support). Pageblocks are processed atomically with the zone +lock held, which can cause severe scheduling delays on both the CPU +going through the pageblock and any other CPUs waiting to acquire the +zone lock. A frequent offender is move_freepages_block(), which is used +by rmqueue() for page allocation. + +As it turns out, there's no requirement for pageblocks to be so large, +so the pageblock order can simply be reduced to ease the scheduling +delays and zone lock contention. PAGE_ALLOC_COSTLY_ORDER is used as a +reasonable setting to ensure non-costly page allocation requests can +still be serviced without always needing to free up more than one +pageblock's worth of pages at a time. + +This has a noticeable effect on overall system latency when memory +pressure is elevated. The various mm functions which operate on +pageblocks no longer appear in the preemptoff tracer, where previously +they would spend up to 100 ms on a mobile arm64 CPU processing a +pageblock with preemption disabled and the zone lock held. + +Signed-off-by: Sultan Alsawaf +--- + include/linux/pageblock-flags.h | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/include/linux/pageblock-flags.h b/include/linux/pageblock-flags.h +index 5f1ae07d724b88..97cda629c9e909 100644 +--- a/include/linux/pageblock-flags.h ++++ b/include/linux/pageblock-flags.h +@@ -48,7 +48,7 @@ extern unsigned int pageblock_order; + #else /* CONFIG_HUGETLB_PAGE */ + + /* If huge pages are not used, group by MAX_ORDER_NR_PAGES */ +-#define pageblock_order MAX_PAGE_ORDER ++#define pageblock_order PAGE_ALLOC_COSTLY_ORDER + + #endif /* CONFIG_HUGETLB_PAGE */ + + +From f22bc56be85e69c71c8e36041193856bb8b01525 Mon Sep 17 00:00:00 2001 +From: Sultan Alsawaf +Date: Wed, 20 Oct 2021 20:50:32 -0700 +Subject: [PATCH] ZEN: mm: Don't hog the CPU and zone lock in rmqueue_bulk() + +There is noticeable scheduling latency and heavy zone lock contention +stemming from rmqueue_bulk's single hold of the zone lock while doing +its work, as seen with the preemptoff tracer. There's no actual need for +rmqueue_bulk() to hold the zone lock the entire time; it only does so +for supposed efficiency. As such, we can relax the zone lock and even +reschedule when IRQs are enabled in order to keep the scheduling delays +and zone lock contention at bay. Forward progress is still guaranteed, +as the zone lock can only be relaxed after page removal. + +With this change, rmqueue_bulk() no longer appears as a serious offender +in the preemptoff tracer, and system latency is noticeably improved. + +Signed-off-by: Sultan Alsawaf +--- + mm/page_alloc.c | 23 ++++++++++++++++++----- + 1 file changed, 18 insertions(+), 5 deletions(-) + +diff --git a/mm/page_alloc.c b/mm/page_alloc.c +index a0b0397e29ee4c..87a983a356530c 100644 +--- a/mm/page_alloc.c ++++ b/mm/page_alloc.c +@@ -3118,15 +3119,16 @@ __rmqueue(struct zone *zone, unsigned int order, int migratetype, + } + + /* +- * Obtain a specified number of elements from the buddy allocator, all under +- * a single hold of the lock, for efficiency. Add them to the supplied list. +- * Returns the number of new pages which were placed at *list. ++ * Obtain a specified number of elements from the buddy allocator, and relax the ++ * zone lock when needed. Add them to the supplied list. Returns the number of ++ * new pages which were placed at *list. + */ + static int rmqueue_bulk(struct zone *zone, unsigned int order, + unsigned long count, struct list_head *list, + int migratetype, unsigned int alloc_flags) + { + unsigned long flags; +- int i; ++ const bool can_resched = !preempt_count() && !irqs_disabled(); ++ int i, allocated = 0, last_mod = 0; + + /* Caller must hold IRQ-safe pcp->lock so IRQs are disabled. */ + spin_lock(&zone->lock); +@@ -3137,6 +3138,18 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order, + if (unlikely(page == NULL)) + break; + ++ /* Reschedule and ease the contention on the lock if needed */ ++ if (i + 1 < count && ((can_resched && need_resched()) || ++ spin_needbreak(&zone->lock))) { ++ __mod_zone_page_state(zone, NR_FREE_PAGES, ++ -((i + 1 - last_mod) << order)); ++ last_mod = i + 1; ++ spin_unlock(&zone->lock); ++ if (can_resched) ++ cond_resched(); ++ spin_lock(&zone->lock); ++ } ++ + if (unlikely(check_pcp_refill(page, order))) + continue; + +@@ -3163,7 +3176,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order, + * on i. Do not confuse with 'allocated' which is the number of + * pages added to the pcp list. + */ +- __mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order)); ++ __mod_zone_page_state(zone, NR_FREE_PAGES, -((i - last_mod) << order)); + spin_unlock(&zone->lock); + return allocated; + } + +From 6329525a0fa10cd13f39b76948b1296150f75c95 Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 29 Aug 2022 16:47:26 +0000 +Subject: [PATCH 14/16] XANMOD: Makefile: Disable GCC vectorization on trees + +Signed-off-by: Alexandre Frade +--- + Makefile | 3 +++ + 1 file changed, 3 insertions(+) + +diff --git a/Makefile b/Makefile +index 3f6628780eb2..35a5ae1ede42 100644 +--- a/Makefile ++++ b/Makefile +@@ -1069,6 +1069,9 @@ endif + KBUILD_CFLAGS-$(call gcc-min-version, 90100) += -Wno-alloc-size-larger-than + KBUILD_CFLAGS += $(KBUILD_CFLAGS-y) $(CONFIG_CC_IMPLICIT_FALLTHROUGH) + ++# disable GCC vectorization on trees ++KBUILD_CFLAGS += $(call cc-option, -fno-tree-vectorize) ++ + # disable invalid "can't wrap" optimizations for signed / pointers + KBUILD_CFLAGS += -fno-strict-overflow + +-- +2.39.1 + +From f997578464b2c4c63e7bd1afbfef56212ee44f2d Mon Sep 17 00:00:00 2001 +From: Etienne JUVIGNY +Date: Mon, 6 Mar 2023 13:54:09 +0100 +Subject: Don't add -dirty versioning on unclean trees + + +diff --git a/scripts/setlocalversion b/scripts/setlocalversion +index ca5795e16..ad0d94477 100755 +--- a/scripts/setlocalversion ++++ b/scripts/setlocalversion +@@ -85,12 +85,12 @@ scm_version() + # git-diff-index does not refresh the index, so it may give misleading + # results. + # See git-update-index(1), git-diff-index(1), and git-status(1). +- if { +- git --no-optional-locks status -uno --porcelain 2>/dev/null || +- git diff-index --name-only HEAD +- } | read dummy; then +- printf '%s' -dirty +- fi ++ #if { ++ # git --no-optional-locks status -uno --porcelain 2>/dev/null || ++ # git diff-index --name-only HEAD ++ #} | read dummy; then ++ # printf '%s' -dirty ++ #fi + } + + collect_files() + +From 1cf70fdd26245554ab30234722338d8160dff394 Mon Sep 17 00:00:00 2001 +From: Steven Barrett +Date: Sat, 21 May 2022 15:15:09 -0500 +Subject: [PATCH] ZEN: INTERACTIVE: dm-crypt: Disable workqueues for crypto ops + +Queueing in dm-crypt for crypto operations reduces performance on modern +systems. As discussed in an article from Cloudflare, they discovered +that queuing was introduced because the crypto subsystem used to be +synchronous. Since it's now asynchronous, we get double queueing when +using the subsystem through dm-crypt. This is obviously undesirable and +reduces throughput and increases latency. + +Disable queueing when using our Zen Interactive configuration. + +Fixes: https://github.com/zen-kernel/zen-kernel/issues/282 + +tkg: Config switch changed to our local "ZENIFY" toggle +--- + drivers/md/dm-crypt.c | 5 +++++ + init/Kconfig | 1 + + 2 files changed, 6 insertions(+) + +diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c +index 2ae8560b6a14ad..cb49218030c88b 100644 +--- a/drivers/md/dm-crypt.c ++++ b/drivers/md/dm-crypt.c +@@ -3242,6 +3242,11 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv) + goto bad; + } + ++#ifdef CONFIG_ZENIFY ++ set_bit(DM_CRYPT_NO_READ_WORKQUEUE, &cc->flags); ++ set_bit(DM_CRYPT_NO_WRITE_WORKQUEUE, &cc->flags); ++#endif ++ + ret = crypt_ctr_cipher(ti, argv[0], argv[1]); + if (ret < 0) + goto bad; diff --git a/patches/sys-kernel/gentoo-sources/0003-glitched-cfs.patch b/patches/sys-kernel/gentoo-sources/0003-glitched-cfs.patch new file mode 100644 index 0000000..f0e4822 --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0003-glitched-cfs.patch @@ -0,0 +1,117 @@ +diff --git a/kernel/Kconfig.hz b/kernel/Kconfig.hz +index 2a202a846757..1d9c7ed79b11 100644 +--- a/kernel/Kconfig.hz ++++ b/kernel/Kconfig.hz +@@ -4,7 +4,7 @@ + + choice + prompt "Timer frequency" +- default HZ_250 ++ default HZ_500 + help + Allows the configuration of the timer frequency. It is customary + to have the timer interrupt run at 1000 Hz but 100 Hz may be more +@@ -39,6 +39,13 @@ choice + on SMP and NUMA systems and exactly dividing by both PAL and + NTSC frame rates for video and multimedia work. + ++ config HZ_500 ++ bool "500 HZ" ++ help ++ 500 Hz is a balanced timer frequency. Provides fast interactivity ++ on desktops with great smoothness without increasing CPU power ++ consumption and sacrificing the battery life on laptops. ++ + config HZ_1000 + bool "1000 HZ" + help +@@ -52,6 +59,7 @@ config HZ + default 100 if HZ_100 + default 250 if HZ_250 + default 300 if HZ_300 ++ default 500 if HZ_500 + default 1000 if HZ_1000 + + config SCHED_HRTICK + +diff --git a/kernel/Kconfig.hz b/kernel/Kconfig.hz +index 2a202a846757..1d9c7ed79b11 100644 +--- a/kernel/Kconfig.hz ++++ b/kernel/Kconfig.hz +@@ -4,7 +4,7 @@ + + choice + prompt "Timer frequency" +- default HZ_500 ++ default HZ_750 + help + Allows the configuration of the timer frequency. It is customary + to have the timer interrupt run at 1000 Hz but 100 Hz may be more +@@ -46,6 +46,13 @@ choice + on desktops with great smoothness without increasing CPU power + consumption and sacrificing the battery life on laptops. + ++ config HZ_750 ++ bool "750 HZ" ++ help ++ 750 Hz is a good timer frequency for desktops. Provides fast ++ interactivity with great smoothness without sacrificing too ++ much throughput. ++ + config HZ_1000 + bool "1000 HZ" + help +@@ -60,6 +67,7 @@ config HZ + default 250 if HZ_250 + default 300 if HZ_300 + default 500 if HZ_500 ++ default 750 if HZ_750 + default 1000 if HZ_1000 + + config SCHED_HRTICK + +diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c +index 6b423eebfd5d..61e3271675d6 100644 +--- a/drivers/cpufreq/cpufreq_ondemand.c ++++ b/drivers/cpufreq/cpufreq_ondemand.c +@@ -21,10 +21,10 @@ + #include "cpufreq_ondemand.h" + + /* On-demand governor macros */ +-#define DEF_FREQUENCY_UP_THRESHOLD (80) +-#define DEF_SAMPLING_DOWN_FACTOR (1) ++#define DEF_FREQUENCY_UP_THRESHOLD (55) ++#define DEF_SAMPLING_DOWN_FACTOR (5) + #define MAX_SAMPLING_DOWN_FACTOR (100000) +-#define MICRO_FREQUENCY_UP_THRESHOLD (95) ++#define MICRO_FREQUENCY_UP_THRESHOLD (63) + #define MIN_FREQUENCY_UP_THRESHOLD (1) + #define MAX_FREQUENCY_UP_THRESHOLD (100) + +From cba31b19f8c38696b13ba48e0e8b6dbe747d6bae Mon Sep 17 00:00:00 2001 +From: Alexandre Frade +Date: Mon, 29 Jan 2018 17:31:25 +0000 +Subject: [PATCH 10/16] XANMOD: mm/vmscan: vm_swappiness = 30 decreases the + amount of swapping + +Signed-off-by: Alexandre Frade +Signed-off-by: Alexandre Frade +--- + mm/vmscan.c | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/mm/vmscan.c b/mm/vmscan.c +index 5b7b8d4f5297..549684b29418 100644 +--- a/mm/vmscan.c ++++ b/mm/vmscan.c +@@ -190,7 +190,7 @@ struct scan_control { + /* + * From 0 .. 200. Higher means more swappy. + */ +-int vm_swappiness = 60; ++int vm_swappiness = 30; + + static void set_task_reclaim_state(struct task_struct *task, + struct reclaim_state *rs) +-- +2.39.1 diff --git a/patches/sys-kernel/gentoo-sources/0005-ksm.patch b/patches/sys-kernel/gentoo-sources/0005-ksm.patch new file mode 100644 index 0000000..edc38c9 --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0005-ksm.patch @@ -0,0 +1,433 @@ +From 3d95a728e5570d4aedad19d699fd23db695f1ada Mon Sep 17 00:00:00 2001 +From: Peter Jung +Date: Mon, 7 Oct 2024 11:53:40 +0200 +Subject: [PATCH 5/8] ksm + +Signed-off-by: Peter Jung +--- + arch/alpha/kernel/syscalls/syscall.tbl | 3 + + arch/arm/tools/syscall.tbl | 3 + + arch/m68k/kernel/syscalls/syscall.tbl | 3 + + arch/microblaze/kernel/syscalls/syscall.tbl | 3 + + arch/mips/kernel/syscalls/syscall_n32.tbl | 3 + + arch/mips/kernel/syscalls/syscall_n64.tbl | 3 + + arch/mips/kernel/syscalls/syscall_o32.tbl | 3 + + arch/parisc/kernel/syscalls/syscall.tbl | 3 + + arch/powerpc/kernel/syscalls/syscall.tbl | 3 + + arch/s390/kernel/syscalls/syscall.tbl | 3 + + arch/sh/kernel/syscalls/syscall.tbl | 3 + + arch/sparc/kernel/syscalls/syscall.tbl | 3 + + arch/x86/entry/syscalls/syscall_32.tbl | 3 + + arch/x86/entry/syscalls/syscall_64.tbl | 3 + + arch/xtensa/kernel/syscalls/syscall.tbl | 3 + + include/linux/syscalls.h | 3 + + include/uapi/asm-generic/unistd.h | 9 +- + kernel/sys.c | 138 ++++++++++++++++++ + kernel/sys_ni.c | 3 + + scripts/syscall.tbl | 3 + + .../arch/powerpc/entry/syscalls/syscall.tbl | 3 + + .../perf/arch/s390/entry/syscalls/syscall.tbl | 3 + + 22 files changed, 206 insertions(+), 1 deletion(-) + +diff --git a/arch/alpha/kernel/syscalls/syscall.tbl b/arch/alpha/kernel/syscalls/syscall.tbl +index 74720667fe09..e6a11f3c0a2e 100644 +--- a/arch/alpha/kernel/syscalls/syscall.tbl ++++ b/arch/alpha/kernel/syscalls/syscall.tbl +@@ -502,3 +502,6 @@ + 570 common lsm_set_self_attr sys_lsm_set_self_attr + 571 common lsm_list_modules sys_lsm_list_modules + 572 common mseal sys_mseal ++573 common process_ksm_enable sys_process_ksm_enable ++574 common process_ksm_disable sys_process_ksm_disable ++575 common process_ksm_status sys_process_ksm_status +diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl +index 23c98203c40f..10a3099decbe 100644 +--- a/arch/arm/tools/syscall.tbl ++++ b/arch/arm/tools/syscall.tbl +@@ -477,3 +477,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/m68k/kernel/syscalls/syscall.tbl b/arch/m68k/kernel/syscalls/syscall.tbl +index 22a3cbd4c602..12d2c7594bf0 100644 +--- a/arch/m68k/kernel/syscalls/syscall.tbl ++++ b/arch/m68k/kernel/syscalls/syscall.tbl +@@ -462,3 +462,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/microblaze/kernel/syscalls/syscall.tbl b/arch/microblaze/kernel/syscalls/syscall.tbl +index 2b81a6bd78b2..e2a93c856eed 100644 +--- a/arch/microblaze/kernel/syscalls/syscall.tbl ++++ b/arch/microblaze/kernel/syscalls/syscall.tbl +@@ -468,3 +468,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/mips/kernel/syscalls/syscall_n32.tbl b/arch/mips/kernel/syscalls/syscall_n32.tbl +index 953f5b7dc723..b921fbf56fa6 100644 +--- a/arch/mips/kernel/syscalls/syscall_n32.tbl ++++ b/arch/mips/kernel/syscalls/syscall_n32.tbl +@@ -401,3 +401,6 @@ + 460 n32 lsm_set_self_attr sys_lsm_set_self_attr + 461 n32 lsm_list_modules sys_lsm_list_modules + 462 n32 mseal sys_mseal ++463 n32 process_ksm_enable sys_process_ksm_enable ++464 n32 process_ksm_disable sys_process_ksm_disable ++465 n32 process_ksm_status sys_process_ksm_status +diff --git a/arch/mips/kernel/syscalls/syscall_n64.tbl b/arch/mips/kernel/syscalls/syscall_n64.tbl +index 1464c6be6eb3..8d7f9ddd66f4 100644 +--- a/arch/mips/kernel/syscalls/syscall_n64.tbl ++++ b/arch/mips/kernel/syscalls/syscall_n64.tbl +@@ -377,3 +377,6 @@ + 460 n64 lsm_set_self_attr sys_lsm_set_self_attr + 461 n64 lsm_list_modules sys_lsm_list_modules + 462 n64 mseal sys_mseal ++463 n64 process_ksm_enable sys_process_ksm_enable ++464 n64 process_ksm_disable sys_process_ksm_disable ++465 n64 process_ksm_status sys_process_ksm_status +diff --git a/arch/mips/kernel/syscalls/syscall_o32.tbl b/arch/mips/kernel/syscalls/syscall_o32.tbl +index 2439a2491cff..9d6142739954 100644 +--- a/arch/mips/kernel/syscalls/syscall_o32.tbl ++++ b/arch/mips/kernel/syscalls/syscall_o32.tbl +@@ -450,3 +450,6 @@ + 460 o32 lsm_set_self_attr sys_lsm_set_self_attr + 461 o32 lsm_list_modules sys_lsm_list_modules + 462 o32 mseal sys_mseal ++463 o32 process_ksm_enable sys_process_ksm_enable ++464 o32 process_ksm_disable sys_process_ksm_disable ++465 o32 process_ksm_status sys_process_ksm_status +diff --git a/arch/parisc/kernel/syscalls/syscall.tbl b/arch/parisc/kernel/syscalls/syscall.tbl +index 66dc406b12e4..9d46476fd908 100644 +--- a/arch/parisc/kernel/syscalls/syscall.tbl ++++ b/arch/parisc/kernel/syscalls/syscall.tbl +@@ -461,3 +461,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/powerpc/kernel/syscalls/syscall.tbl b/arch/powerpc/kernel/syscalls/syscall.tbl +index ebae8415dfbb..16f71bc2f6f0 100644 +--- a/arch/powerpc/kernel/syscalls/syscall.tbl ++++ b/arch/powerpc/kernel/syscalls/syscall.tbl +@@ -553,3 +553,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/s390/kernel/syscalls/syscall.tbl b/arch/s390/kernel/syscalls/syscall.tbl +index 01071182763e..7394bad8178e 100644 +--- a/arch/s390/kernel/syscalls/syscall.tbl ++++ b/arch/s390/kernel/syscalls/syscall.tbl +@@ -465,3 +465,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status sys_process_ksm_status +diff --git a/arch/sh/kernel/syscalls/syscall.tbl b/arch/sh/kernel/syscalls/syscall.tbl +index c55fd7696d40..b9fc31221b87 100644 +--- a/arch/sh/kernel/syscalls/syscall.tbl ++++ b/arch/sh/kernel/syscalls/syscall.tbl +@@ -466,3 +466,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/sparc/kernel/syscalls/syscall.tbl b/arch/sparc/kernel/syscalls/syscall.tbl +index cfdfb3707c16..0d79fd772854 100644 +--- a/arch/sparc/kernel/syscalls/syscall.tbl ++++ b/arch/sparc/kernel/syscalls/syscall.tbl +@@ -508,3 +508,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl +index 534c74b14fab..c546a30575f1 100644 +--- a/arch/x86/entry/syscalls/syscall_32.tbl ++++ b/arch/x86/entry/syscalls/syscall_32.tbl +@@ -468,3 +468,6 @@ + 460 i386 lsm_set_self_attr sys_lsm_set_self_attr + 461 i386 lsm_list_modules sys_lsm_list_modules + 462 i386 mseal sys_mseal ++463 i386 process_ksm_enable sys_process_ksm_enable ++464 i386 process_ksm_disable sys_process_ksm_disable ++465 i386 process_ksm_status sys_process_ksm_status +diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl +index 7093ee21c0d1..0fcd10ba8dfe 100644 +--- a/arch/x86/entry/syscalls/syscall_64.tbl ++++ b/arch/x86/entry/syscalls/syscall_64.tbl +@@ -386,6 +386,9 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status + + # + # Due to a historical design error, certain syscalls are numbered differently +diff --git a/arch/xtensa/kernel/syscalls/syscall.tbl b/arch/xtensa/kernel/syscalls/syscall.tbl +index 67083fc1b2f5..c1aecee4ad9b 100644 +--- a/arch/xtensa/kernel/syscalls/syscall.tbl ++++ b/arch/xtensa/kernel/syscalls/syscall.tbl +@@ -433,3 +433,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h +index 5758104921e6..cc9c4fac2412 100644 +--- a/include/linux/syscalls.h ++++ b/include/linux/syscalls.h +@@ -818,6 +818,9 @@ asmlinkage long sys_madvise(unsigned long start, size_t len, int behavior); + asmlinkage long sys_process_madvise(int pidfd, const struct iovec __user *vec, + size_t vlen, int behavior, unsigned int flags); + asmlinkage long sys_process_mrelease(int pidfd, unsigned int flags); ++asmlinkage long sys_process_ksm_enable(int pidfd, unsigned int flags); ++asmlinkage long sys_process_ksm_disable(int pidfd, unsigned int flags); ++asmlinkage long sys_process_ksm_status(int pidfd, unsigned int flags); + asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size, + unsigned long prot, unsigned long pgoff, + unsigned long flags); +diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h +index 5bf6148cac2b..613e559ad6e0 100644 +--- a/include/uapi/asm-generic/unistd.h ++++ b/include/uapi/asm-generic/unistd.h +@@ -841,8 +841,15 @@ __SYSCALL(__NR_lsm_list_modules, sys_lsm_list_modules) + #define __NR_mseal 462 + __SYSCALL(__NR_mseal, sys_mseal) + ++#define __NR_process_ksm_enable 463 ++__SYSCALL(__NR_process_ksm_enable, sys_process_ksm_enable) ++#define __NR_process_ksm_disable 464 ++__SYSCALL(__NR_process_ksm_disable, sys_process_ksm_disable) ++#define __NR_process_ksm_status 465 ++__SYSCALL(__NR_process_ksm_status, sys_process_ksm_status) ++ + #undef __NR_syscalls +-#define __NR_syscalls 463 ++#define __NR_syscalls 466 + + /* + * 32 bit systems traditionally used different +diff --git a/kernel/sys.c b/kernel/sys.c +index 4da31f28fda8..fcd3aeaddd05 100644 +--- a/kernel/sys.c ++++ b/kernel/sys.c +@@ -2791,6 +2791,144 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, + return error; + } + ++#ifdef CONFIG_KSM ++enum pkc_action { ++ PKSM_ENABLE = 0, ++ PKSM_DISABLE, ++ PKSM_STATUS, ++}; ++ ++static long do_process_ksm_control(int pidfd, enum pkc_action action) ++{ ++ long ret; ++ struct task_struct *task; ++ struct mm_struct *mm; ++ unsigned int f_flags; ++ ++ task = pidfd_get_task(pidfd, &f_flags); ++ if (IS_ERR(task)) { ++ ret = PTR_ERR(task); ++ goto out; ++ } ++ ++ /* Require PTRACE_MODE_READ to avoid leaking ASLR metadata. */ ++ mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); ++ if (IS_ERR_OR_NULL(mm)) { ++ ret = IS_ERR(mm) ? PTR_ERR(mm) : -ESRCH; ++ goto release_task; ++ } ++ ++ /* Require CAP_SYS_NICE for influencing process performance. */ ++ if (!capable(CAP_SYS_NICE)) { ++ ret = -EPERM; ++ goto release_mm; ++ } ++ ++ if (mmap_write_lock_killable(mm)) { ++ ret = -EINTR; ++ goto release_mm; ++ } ++ ++ switch (action) { ++ case PKSM_ENABLE: ++ ret = ksm_enable_merge_any(mm); ++ break; ++ case PKSM_DISABLE: ++ ret = ksm_disable_merge_any(mm); ++ break; ++ case PKSM_STATUS: ++ ret = !!test_bit(MMF_VM_MERGE_ANY, &mm->flags); ++ break; ++ } ++ ++ mmap_write_unlock(mm); ++ ++release_mm: ++ mmput(mm); ++release_task: ++ put_task_struct(task); ++out: ++ return ret; ++} ++#endif /* CONFIG_KSM */ ++ ++SYSCALL_DEFINE2(process_ksm_enable, int, pidfd, unsigned int, flags) ++{ ++#ifdef CONFIG_KSM ++ if (flags != 0) ++ return -EINVAL; ++ ++ return do_process_ksm_control(pidfd, PKSM_ENABLE); ++#else /* CONFIG_KSM */ ++ return -ENOSYS; ++#endif /* CONFIG_KSM */ ++} ++ ++SYSCALL_DEFINE2(process_ksm_disable, int, pidfd, unsigned int, flags) ++{ ++#ifdef CONFIG_KSM ++ if (flags != 0) ++ return -EINVAL; ++ ++ return do_process_ksm_control(pidfd, PKSM_DISABLE); ++#else /* CONFIG_KSM */ ++ return -ENOSYS; ++#endif /* CONFIG_KSM */ ++} ++ ++SYSCALL_DEFINE2(process_ksm_status, int, pidfd, unsigned int, flags) ++{ ++#ifdef CONFIG_KSM ++ if (flags != 0) ++ return -EINVAL; ++ ++ return do_process_ksm_control(pidfd, PKSM_STATUS); ++#else /* CONFIG_KSM */ ++ return -ENOSYS; ++#endif /* CONFIG_KSM */ ++} ++ ++#ifdef CONFIG_KSM ++static ssize_t process_ksm_enable_show(struct kobject *kobj, ++ struct kobj_attribute *attr, char *buf) ++{ ++ return sprintf(buf, "%u\n", __NR_process_ksm_enable); ++} ++static struct kobj_attribute process_ksm_enable_attr = __ATTR_RO(process_ksm_enable); ++ ++static ssize_t process_ksm_disable_show(struct kobject *kobj, ++ struct kobj_attribute *attr, char *buf) ++{ ++ return sprintf(buf, "%u\n", __NR_process_ksm_disable); ++} ++static struct kobj_attribute process_ksm_disable_attr = __ATTR_RO(process_ksm_disable); ++ ++static ssize_t process_ksm_status_show(struct kobject *kobj, ++ struct kobj_attribute *attr, char *buf) ++{ ++ return sprintf(buf, "%u\n", __NR_process_ksm_status); ++} ++static struct kobj_attribute process_ksm_status_attr = __ATTR_RO(process_ksm_status); ++ ++static struct attribute *process_ksm_sysfs_attrs[] = { ++ &process_ksm_enable_attr.attr, ++ &process_ksm_disable_attr.attr, ++ &process_ksm_status_attr.attr, ++ NULL, ++}; ++ ++static const struct attribute_group process_ksm_sysfs_attr_group = { ++ .attrs = process_ksm_sysfs_attrs, ++ .name = "process_ksm", ++}; ++ ++static int __init process_ksm_sysfs_init(void) ++{ ++ return sysfs_create_group(kernel_kobj, &process_ksm_sysfs_attr_group); ++} ++subsys_initcall(process_ksm_sysfs_init); ++#endif /* CONFIG_KSM */ ++ + SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, + struct getcpu_cache __user *, unused) + { +diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c +index c00a86931f8c..d82213d68522 100644 +--- a/kernel/sys_ni.c ++++ b/kernel/sys_ni.c +@@ -186,6 +186,9 @@ COND_SYSCALL(mincore); + COND_SYSCALL(madvise); + COND_SYSCALL(process_madvise); + COND_SYSCALL(process_mrelease); ++COND_SYSCALL(process_ksm_enable); ++COND_SYSCALL(process_ksm_disable); ++COND_SYSCALL(process_ksm_status); + COND_SYSCALL(remap_file_pages); + COND_SYSCALL(mbind); + COND_SYSCALL(get_mempolicy); +diff --git a/scripts/syscall.tbl b/scripts/syscall.tbl +index 845e24eb372e..227d9cc12365 100644 +--- a/scripts/syscall.tbl ++++ b/scripts/syscall.tbl +@@ -403,3 +403,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl b/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl +index ebae8415dfbb..16f71bc2f6f0 100644 +--- a/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl ++++ b/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl +@@ -553,3 +553,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status +diff --git a/tools/perf/arch/s390/entry/syscalls/syscall.tbl b/tools/perf/arch/s390/entry/syscalls/syscall.tbl +index 01071182763e..7394bad8178e 100644 +--- a/tools/perf/arch/s390/entry/syscalls/syscall.tbl ++++ b/tools/perf/arch/s390/entry/syscalls/syscall.tbl +@@ -465,3 +465,6 @@ + 460 common lsm_set_self_attr sys_lsm_set_self_attr sys_lsm_set_self_attr + 461 common lsm_list_modules sys_lsm_list_modules sys_lsm_list_modules + 462 common mseal sys_mseal sys_mseal ++463 common process_ksm_enable sys_process_ksm_enable sys_process_ksm_enable ++464 common process_ksm_disable sys_process_ksm_disable sys_process_ksm_disable ++465 common process_ksm_status sys_process_ksm_status sys_process_ksm_status +-- +2.47.0.rc0 + diff --git a/patches/sys-kernel/gentoo-sources/0006-add-acs-overrides_iommu.patch b/patches/sys-kernel/gentoo-sources/0006-add-acs-overrides_iommu.patch new file mode 100644 index 0000000..562c51f --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0006-add-acs-overrides_iommu.patch @@ -0,0 +1,193 @@ +From cdeab384f48dd9c88e2dff2e9ad8d57dca1a1b1c Mon Sep 17 00:00:00 2001 +From: Mark Weiman +Date: Sun, 12 Aug 2018 11:36:21 -0400 +Subject: [PATCH] pci: Enable overrides for missing ACS capabilities + +This an updated version of Alex Williamson's patch from: +https://lkml.org/lkml/2013/5/30/513 + +Original commit message follows: + +PCIe ACS (Access Control Services) is the PCIe 2.0+ feature that +allows us to control whether transactions are allowed to be redirected +in various subnodes of a PCIe topology. For instance, if two +endpoints are below a root port or downsteam switch port, the +downstream port may optionally redirect transactions between the +devices, bypassing upstream devices. The same can happen internally +on multifunction devices. The transaction may never be visible to the +upstream devices. + +One upstream device that we particularly care about is the IOMMU. If +a redirection occurs in the topology below the IOMMU, then the IOMMU +cannot provide isolation between devices. This is why the PCIe spec +encourages topologies to include ACS support. Without it, we have to +assume peer-to-peer DMA within a hierarchy can bypass IOMMU isolation. + +Unfortunately, far too many topologies do not support ACS to make this +a steadfast requirement. Even the latest chipsets from Intel are only +sporadically supporting ACS. We have trouble getting interconnect +vendors to include the PCIe spec required PCIe capability, let alone +suggested features. + +Therefore, we need to add some flexibility. The pcie_acs_override= +boot option lets users opt-in specific devices or sets of devices to +assume ACS support. The "downstream" option assumes full ACS support +on root ports and downstream switch ports. The "multifunction" +option assumes the subset of ACS features available on multifunction +endpoints and upstream switch ports are supported. The "id:nnnn:nnnn" +option enables ACS support on devices matching the provided vendor +and device IDs, allowing more strategic ACS overrides. These options +may be combined in any order. A maximum of 16 id specific overrides +are available. It's suggested to use the most limited set of options +necessary to avoid completely disabling ACS across the topology. +Note to hardware vendors, we have facilities to permanently quirk +specific devices which enforce isolation but not provide an ACS +capability. Please contact me to have your devices added and save +your customers the hassle of this boot option. + +Signed-off-by: Mark Weiman +--- + .../admin-guide/kernel-parameters.txt | 9 ++ + drivers/pci/quirks.c | 101 ++++++++++++++++++ + 2 files changed, 110 insertions(+) + +diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt +index aefd358a5ca3..173b3596fd9e 100644 +--- a/Documentation/admin-guide/kernel-parameters.txt ++++ b/Documentation/admin-guide/kernel-parameters.txt +@@ -3190,6 +3190,15 @@ + nomsi [MSI] If the PCI_MSI kernel config parameter is + enabled, this kernel boot option can be used to + disable the use of MSI interrupts system-wide. ++ pcie_acs_override = ++ [PCIE] Override missing PCIe ACS support for: ++ downstream ++ All downstream ports - full ACS capabilities ++ multifunction ++ All multifunction devices - multifunction ACS subset ++ id:nnnn:nnnn ++ Specific device - full ACS capabilities ++ Specified as vid:did (vendor/device ID) in hex + noioapicquirk [APIC] Disable all boot interrupt quirks. + Safety option to keep boot IRQs enabled. This + should never be necessary. +diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c +index 4700d24e5d55..8f7a3d7fd9c1 100644 +--- a/drivers/pci/quirks.c ++++ b/drivers/pci/quirks.c +@@ -3372,6 +3372,106 @@ static void quirk_no_bus_reset(struct pci_dev *dev) + dev->dev_flags |= PCI_DEV_FLAGS_NO_BUS_RESET; + } + ++static bool acs_on_downstream; ++static bool acs_on_multifunction; ++ ++#define NUM_ACS_IDS 16 ++struct acs_on_id { ++ unsigned short vendor; ++ unsigned short device; ++}; ++static struct acs_on_id acs_on_ids[NUM_ACS_IDS]; ++static u8 max_acs_id; ++ ++static __init int pcie_acs_override_setup(char *p) ++{ ++ if (!p) ++ return -EINVAL; ++ ++ while (*p) { ++ if (!strncmp(p, "downstream", 10)) ++ acs_on_downstream = true; ++ if (!strncmp(p, "multifunction", 13)) ++ acs_on_multifunction = true; ++ if (!strncmp(p, "id:", 3)) { ++ char opt[5]; ++ int ret; ++ long val; ++ ++ if (max_acs_id >= NUM_ACS_IDS - 1) { ++ pr_warn("Out of PCIe ACS override slots (%d)\n", ++ NUM_ACS_IDS); ++ goto next; ++ } ++ ++ p += 3; ++ snprintf(opt, 5, "%s", p); ++ ret = kstrtol(opt, 16, &val); ++ if (ret) { ++ pr_warn("PCIe ACS ID parse error %d\n", ret); ++ goto next; ++ } ++ acs_on_ids[max_acs_id].vendor = val; ++ ++ p += strcspn(p, ":"); ++ if (*p != ':') { ++ pr_warn("PCIe ACS invalid ID\n"); ++ goto next; ++ } ++ ++ p++; ++ snprintf(opt, 5, "%s", p); ++ ret = kstrtol(opt, 16, &val); ++ if (ret) { ++ pr_warn("PCIe ACS ID parse error %d\n", ret); ++ goto next; ++ } ++ acs_on_ids[max_acs_id].device = val; ++ max_acs_id++; ++ } ++next: ++ p += strcspn(p, ","); ++ if (*p == ',') ++ p++; ++ } ++ ++ if (acs_on_downstream || acs_on_multifunction || max_acs_id) ++ pr_warn("Warning: PCIe ACS overrides enabled; This may allow non-IOMMU protected peer-to-peer DMA\n"); ++ ++ return 0; ++} ++early_param("pcie_acs_override", pcie_acs_override_setup); ++ ++static int pcie_acs_overrides(struct pci_dev *dev, u16 acs_flags) ++{ ++ int i; ++ ++ /* Never override ACS for legacy devices or devices with ACS caps */ ++ if (!pci_is_pcie(dev) || ++ pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS)) ++ return -ENOTTY; ++ ++ for (i = 0; i < max_acs_id; i++) ++ if (acs_on_ids[i].vendor == dev->vendor && ++ acs_on_ids[i].device == dev->device) ++ return 1; ++ ++ switch (pci_pcie_type(dev)) { ++ case PCI_EXP_TYPE_DOWNSTREAM: ++ case PCI_EXP_TYPE_ROOT_PORT: ++ if (acs_on_downstream) ++ return 1; ++ break; ++ case PCI_EXP_TYPE_ENDPOINT: ++ case PCI_EXP_TYPE_UPSTREAM: ++ case PCI_EXP_TYPE_LEG_END: ++ case PCI_EXP_TYPE_RC_END: ++ if (acs_on_multifunction && dev->multifunction) ++ return 1; ++ } ++ ++ return -ENOTTY; ++} + /* + * Some Atheros AR9xxx and QCA988x chips do not behave after a bus reset. + * The device will throw a Link Down error on AER-capable systems and +@@ -5102,6 +5102,7 @@ + { PCI_VENDOR_ID_ZHAOXIN, PCI_ANY_ID, pci_quirk_zhaoxin_pcie_ports_acs }, + /* Wangxun nics */ + { PCI_VENDOR_ID_WANGXUN, PCI_ANY_ID, pci_quirk_wangxun_nic_acs }, ++ { PCI_ANY_ID, PCI_ANY_ID, pcie_acs_overrides }, + { 0 } + }; + + diff --git a/patches/sys-kernel/gentoo-sources/0007-v6.12-fsync_legacy_via_futex_waitv.patch b/patches/sys-kernel/gentoo-sources/0007-v6.12-fsync_legacy_via_futex_waitv.patch new file mode 100644 index 0000000..fdda084 --- /dev/null +++ b/patches/sys-kernel/gentoo-sources/0007-v6.12-fsync_legacy_via_futex_waitv.patch @@ -0,0 +1,166 @@ +From b70e738f08403950aa3053c36b98c6b0eeb0eb90 Mon Sep 17 00:00:00 2001 +From: =?UTF-8?q?Andr=C3=A9=20Almeida?= +Date: Mon, 25 Oct 2021 09:49:42 -0300 +Subject: [PATCH] futex: Add entry point for FUTEX_WAIT_MULTIPLE (opcode 31) +MIME-Version: 1.0 +Content-Type: text/plain; charset=UTF-8 +Content-Transfer-Encoding: 8bit + +Add an option to wait on multiple futexes using the old interface, that +uses opcode 31 through futex() syscall. Do that by just translation the +old interface to use the new code. This allows old and stable versions +of Proton to still use fsync in new kernel releases. + +Signed-off-by: AndrĂ© Almeida +--- + include/uapi/linux/futex.h | 13 +++++++ + kernel/futex/syscalls.c | 75 +++++++++++++++++++++++++++++++++++++- + 2 files changed, 87 insertions(+), 1 deletion(-) + +diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h +index 71a5df8d2689..d375ab21cbf8 100644 +--- a/include/uapi/linux/futex.h ++++ b/include/uapi/linux/futex.h +@@ -22,6 +22,7 @@ + #define FUTEX_WAIT_REQUEUE_PI 11 + #define FUTEX_CMP_REQUEUE_PI 12 + #define FUTEX_LOCK_PI2 13 ++#define FUTEX_WAIT_MULTIPLE 31 + + #define FUTEX_PRIVATE_FLAG 128 + #define FUTEX_CLOCK_REALTIME 256 +@@ -68,6 +69,18 @@ struct futex_waitv { + __u32 __reserved; + }; + ++/** ++ * struct futex_wait_block - Block of futexes to be waited for ++ * @uaddr: User address of the futex ++ * @val: Futex value expected by userspace ++ * @bitset: Bitset for the optional bitmasked wakeup ++ */ ++struct futex_wait_block { ++ __u32 __user *uaddr; ++ __u32 val; ++ __u32 bitset; ++}; ++ + /* + * Support for robust futexes: the kernel cleans up held futexes at + * thread exit time. +diff --git a/kernel/futex/syscalls.c b/kernel/futex/syscalls.c +index 6f91a07a6a83..2f4d4c04ede2 100644 +--- a/kernel/futex/syscalls.c ++++ b/kernel/futex/syscalls.c +@@ -158,6 +158,7 @@ static __always_inline bool futex_cmd_has_timeout(u32 cmd) + case FUTEX_LOCK_PI2: + case FUTEX_WAIT_BITSET: + case FUTEX_WAIT_REQUEUE_PI: ++ case FUTEX_WAIT_MULTIPLE: + return true; + } + return false; +@@ -170,13 +171,79 @@ futex_init_timeout(u32 cmd, u32 op, struct timespec64 *ts, ktime_t *t) + return -EINVAL; + + *t = timespec64_to_ktime(*ts); +- if (cmd == FUTEX_WAIT) ++ if (cmd == FUTEX_WAIT || cmd == FUTEX_WAIT_MULTIPLE) + *t = ktime_add_safe(ktime_get(), *t); + else if (cmd != FUTEX_LOCK_PI && !(op & FUTEX_CLOCK_REALTIME)) + *t = timens_ktime_to_host(CLOCK_MONOTONIC, *t); + return 0; + } + ++/** ++ * futex_read_wait_block - Read an array of futex_wait_block from userspace ++ * @uaddr: Userspace address of the block ++ * @count: Number of blocks to be read ++ * ++ * This function creates and allocate an array of futex_q (we zero it to ++ * initialize the fields) and then, for each futex_wait_block element from ++ * userspace, fill a futex_q element with proper values. ++ */ ++inline struct futex_vector *futex_read_wait_block(u32 __user *uaddr, u32 count) ++{ ++ unsigned int i; ++ struct futex_vector *futexv; ++ struct futex_wait_block fwb; ++ struct futex_wait_block __user *entry = ++ (struct futex_wait_block __user *)uaddr; ++ ++ if (!count || count > FUTEX_WAITV_MAX) ++ return ERR_PTR(-EINVAL); ++ ++ futexv = kcalloc(count, sizeof(*futexv), GFP_KERNEL); ++ if (!futexv) ++ return ERR_PTR(-ENOMEM); ++ ++ for (i = 0; i < count; i++) { ++ if (copy_from_user(&fwb, &entry[i], sizeof(fwb))) { ++ kfree(futexv); ++ return ERR_PTR(-EFAULT); ++ } ++ ++ futexv[i].w.flags = FUTEX_32; ++ futexv[i].w.val = fwb.val; ++ futexv[i].w.uaddr = (uintptr_t) (fwb.uaddr); ++ futexv[i].q = futex_q_init; ++ } ++ ++ return futexv; ++} ++ ++int futex_wait_multiple(struct futex_vector *vs, unsigned int count, ++ struct hrtimer_sleeper *to); ++ ++int futex_opcode_31(ktime_t *abs_time, u32 __user *uaddr, int count) ++{ ++ int ret; ++ struct futex_vector *vs; ++ struct hrtimer_sleeper *to = NULL, timeout; ++ ++ to = futex_setup_timer(abs_time, &timeout, 0, 0); ++ ++ vs = futex_read_wait_block(uaddr, count); ++ ++ if (IS_ERR(vs)) ++ return PTR_ERR(vs); ++ ++ ret = futex_wait_multiple(vs, count, abs_time ? to : NULL); ++ kfree(vs); ++ ++ if (to) { ++ hrtimer_cancel(&to->timer); ++ destroy_hrtimer_on_stack(&to->timer); ++ } ++ ++ return ret; ++} ++ + SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val, + const struct __kernel_timespec __user *, utime, + u32 __user *, uaddr2, u32, val3) +@@ -196,6 +263,9 @@ SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val, + tp = &t; + } + ++ if (cmd == FUTEX_WAIT_MULTIPLE) ++ return futex_opcode_31(tp, uaddr, val); ++ + return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3); + } + +@@ -392,6 +462,9 @@ SYSCALL_DEFINE6(futex_time32, u32 __user *, uaddr, int, op, u32, val, + tp = &t; + } + ++ if (cmd == FUTEX_WAIT_MULTIPLE) ++ return futex_opcode_31(tp, uaddr, val); ++ + return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3); + } + #endif /* CONFIG_COMPAT_32BIT_TIME */ +-- +2.33.1 + diff --git a/patches/sys-kernel/gentoo-sources/0008-zstd.patch b/patches/sys-kernel/gentoo-sources/0008-zstd.patch index 44c6deb..a32a7bc 100644 --- a/patches/sys-kernel/gentoo-sources/0008-zstd.patch +++ b/patches/sys-kernel/gentoo-sources/0008-zstd.patch @@ -1,6 +1,6 @@ -From ef7086e8c29b16dc397fcf7e29f4608865f41cad Mon Sep 17 00:00:00 2001 +From 9342b92704267791d2b57a24f3a0db84c5f00912 Mon Sep 17 00:00:00 2001 From: Peter Jung -Date: Wed, 3 Apr 2024 17:07:41 +0200 +Date: Mon, 7 Oct 2024 11:54:11 +0200 Subject: [PATCH 8/8] zstd Signed-off-by: Peter Jung @@ -68,7 +68,7 @@ Signed-off-by: Peter Jung create mode 100644 lib/zstd/common/bits.h diff --git a/include/linux/zstd.h b/include/linux/zstd.h -index 113408eef6ec..f109d49f43f8 100644 +index b2c7cf310c8f..ac59ae9a18d7 100644 --- a/include/linux/zstd.h +++ b/include/linux/zstd.h @@ -1,6 +1,6 @@ @@ -3485,7 +3485,7 @@ index 5042ff870308..57462466e188 100644 +#endif /* HUF_H_298734234 */ diff --git a/lib/zstd/common/mem.h b/lib/zstd/common/mem.h -index 1d9cc03924ca..2e91e7780c1f 100644 +index c22a2e69bf46..d9bd752fe17b 100644 --- a/lib/zstd/common/mem.h +++ b/lib/zstd/common/mem.h @@ -1,6 +1,6 @@ @@ -4965,7 +4965,7 @@ index 74ef0db47621..0b229f5d2ae2 100644 } - diff --git a/lib/zstd/compress/zstd_compress.c b/lib/zstd/compress/zstd_compress.c -index f620cafca633..0d139727cd39 100644 +index 16bb995bc6c4..885167f7e47b 100644 --- a/lib/zstd/compress/zstd_compress.c +++ b/lib/zstd/compress/zstd_compress.c @@ -1,5 +1,6 @@ @@ -8017,16 +8017,16 @@ index f620cafca633..0d139727cd39 100644 FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed"); assert(dictID <= (size_t)(U32)-1); cdict->dictID = (U32)dictID; -@@ -4811,7 +5450,7 @@ ZSTD_CDict* ZSTD_createCDict_advanced2( - cctxParams.useRowMatchFinder, cctxParams.enableDedicatedDictSearch, - customMem); +@@ -4813,7 +5452,7 @@ ZSTD_CDict* ZSTD_createCDict_advanced2( + if (!cdict) + return NULL; - if (ZSTD_isError( ZSTD_initCDict_internal(cdict, + if (!cdict || ZSTD_isError( ZSTD_initCDict_internal(cdict, dict, dictSize, dictLoadMethod, dictContentType, cctxParams) )) { -@@ -4906,6 +5545,7 @@ const ZSTD_CDict* ZSTD_initStaticCDict( +@@ -4908,6 +5547,7 @@ const ZSTD_CDict* ZSTD_initStaticCDict( params.cParams = cParams; params.useRowMatchFinder = useRowMatchFinder; cdict->useRowMatchFinder = useRowMatchFinder; @@ -8034,7 +8034,7 @@ index f620cafca633..0d139727cd39 100644 if (ZSTD_isError( ZSTD_initCDict_internal(cdict, dict, dictSize, -@@ -4985,12 +5625,17 @@ size_t ZSTD_compressBegin_usingCDict_advanced( +@@ -4987,12 +5627,17 @@ size_t ZSTD_compressBegin_usingCDict_advanced( /* ZSTD_compressBegin_usingCDict() : * cdict must be != NULL */ @@ -8053,7 +8053,7 @@ index f620cafca633..0d139727cd39 100644 /*! ZSTD_compress_usingCDict_internal(): * Implementation of various ZSTD_compress_usingCDict* functions. */ -@@ -5000,7 +5645,7 @@ static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx, +@@ -5002,7 +5647,7 @@ static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams) { FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */ @@ -8062,7 +8062,7 @@ index f620cafca633..0d139727cd39 100644 } /*! ZSTD_compress_usingCDict_advanced(): -@@ -5197,30 +5842,41 @@ size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel) +@@ -5199,30 +5844,41 @@ size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel) static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx) { @@ -8116,7 +8116,7 @@ index f620cafca633..0d139727cd39 100644 if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) { assert(zcs->inBuff != NULL); assert(zcs->inBuffSize > 0); -@@ -5229,8 +5885,10 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5231,8 +5887,10 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, assert(zcs->outBuff != NULL); assert(zcs->outBuffSize > 0); } @@ -8128,7 +8128,7 @@ index f620cafca633..0d139727cd39 100644 assert((U32)flushMode <= (U32)ZSTD_e_end); while (someMoreWork) { -@@ -5245,7 +5903,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5247,7 +5905,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) /* OR we are allowed to return dstSizeTooSmall */ && (zcs->inBuffPos == 0) ) { /* shortcut to compression pass directly into output buffer */ @@ -8137,7 +8137,7 @@ index f620cafca633..0d139727cd39 100644 op, oend-op, ip, iend-ip); DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize); FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed"); -@@ -5262,8 +5920,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5264,8 +5922,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, zcs->inBuff + zcs->inBuffPos, toLoad, ip, iend-ip); zcs->inBuffPos += loaded; @@ -8147,7 +8147,7 @@ index f620cafca633..0d139727cd39 100644 if ( (flushMode == ZSTD_e_continue) && (zcs->inBuffPos < zcs->inBuffTarget) ) { /* not enough input to fill full block : stop here */ -@@ -5274,6 +5931,20 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5276,6 +5933,20 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, /* empty */ someMoreWork = 0; break; } @@ -8168,7 +8168,7 @@ index f620cafca633..0d139727cd39 100644 } /* compress current block (note : this stage cannot be stopped in the middle) */ DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode); -@@ -5281,9 +5952,8 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5283,9 +5954,8 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, void* cDst; size_t cSize; size_t oSize = oend-op; @@ -8180,7 +8180,7 @@ index f620cafca633..0d139727cd39 100644 if (oSize >= ZSTD_compressBound(iSize) || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) cDst = op; /* compress into output buffer, to skip flush stage */ else -@@ -5291,9 +5961,9 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5293,9 +5963,9 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, if (inputBuffered) { unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend); cSize = lastBlock ? @@ -8192,7 +8192,7 @@ index f620cafca633..0d139727cd39 100644 zcs->inBuff + zcs->inToCompress, iSize); FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed"); zcs->frameEnded = lastBlock; -@@ -5306,19 +5976,16 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, +@@ -5308,19 +5978,16 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, if (!lastBlock) assert(zcs->inBuffTarget <= zcs->inBuffSize); zcs->inToCompress = zcs->inBuffPos; @@ -8218,7 +8218,7 @@ index f620cafca633..0d139727cd39 100644 } if (cDst == op) { /* no need to flush */ op += cSize; -@@ -5388,8 +6055,10 @@ size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuf +@@ -5390,8 +6057,10 @@ size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuf /* After a compression call set the expected input/output buffer. * This is validated at the start of the next compression call. */ @@ -8230,7 +8230,7 @@ index f620cafca633..0d139727cd39 100644 if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { cctx->expectedInBuffer = *input; } -@@ -5408,22 +6077,22 @@ static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx, +@@ -5410,22 +6079,22 @@ static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx, { if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { ZSTD_inBuffer const expect = cctx->expectedInBuffer; @@ -8259,7 +8259,7 @@ index f620cafca633..0d139727cd39 100644 ZSTD_CCtx_params params = cctx->requestedParams; ZSTD_prefixDict const prefixDict = cctx->prefixDict; FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */ -@@ -5437,9 +6106,9 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, +@@ -5439,9 +6108,9 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, params.compressionLevel = cctx->cdict->compressionLevel; } DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage"); @@ -8272,7 +8272,7 @@ index f620cafca633..0d139727cd39 100644 ? prefixDict.dictSize : (cctx->cdict ? cctx->cdict->dictContentSize : 0); ZSTD_cParamMode_e const mode = ZSTD_getCParamMode(cctx->cdict, ¶ms, cctx->pledgedSrcSizePlusOne - 1); -@@ -5451,6 +6120,9 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, +@@ -5453,6 +6122,9 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, params.useBlockSplitter = ZSTD_resolveBlockSplitterMode(params.useBlockSplitter, ¶ms.cParams); params.ldmParams.enableLdm = ZSTD_resolveEnableLdm(params.ldmParams.enableLdm, ¶ms.cParams); params.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params.useRowMatchFinder, ¶ms.cParams); @@ -8282,7 +8282,7 @@ index f620cafca633..0d139727cd39 100644 { U64 const pledgedSrcSize = cctx->pledgedSrcSizePlusOne - 1; assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams))); -@@ -5477,6 +6149,8 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, +@@ -5479,6 +6151,8 @@ static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, return 0; } @@ -8291,7 +8291,7 @@ index f620cafca633..0d139727cd39 100644 size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input, -@@ -5491,8 +6165,27 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, +@@ -5493,8 +6167,27 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, /* transparent initialization stage */ if (cctx->streamStage == zcss_init) { @@ -8321,7 +8321,7 @@ index f620cafca633..0d139727cd39 100644 } /* end of transparent initialization stage */ -@@ -5510,13 +6203,20 @@ size_t ZSTD_compressStream2_simpleArgs ( +@@ -5512,13 +6205,20 @@ size_t ZSTD_compressStream2_simpleArgs ( const void* src, size_t srcSize, size_t* srcPos, ZSTD_EndDirective endOp) { @@ -8348,7 +8348,7 @@ index f620cafca633..0d139727cd39 100644 } size_t ZSTD_compress2(ZSTD_CCtx* cctx, -@@ -5539,6 +6239,7 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx, +@@ -5541,6 +6241,7 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx, /* Reset to the original values. */ cctx->requestedParams.inBufferMode = originalInBufferMode; cctx->requestedParams.outBufferMode = originalOutBufferMode; @@ -8356,7 +8356,7 @@ index f620cafca633..0d139727cd39 100644 FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed"); if (result != 0) { /* compression not completed, due to lack of output space */ assert(oPos == dstCapacity); -@@ -5549,64 +6250,61 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx, +@@ -5551,64 +6252,61 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx, } } @@ -8441,7 +8441,7 @@ index f620cafca633..0d139727cd39 100644 if (cctx->cdict) { dictSize = (U32)cctx->cdict->dictContentSize; } else if (cctx->prefixDict.dict) { -@@ -5615,25 +6313,55 @@ ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, +@@ -5617,25 +6315,55 @@ ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, dictSize = 0; } ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); @@ -8506,7 +8506,7 @@ index f620cafca633..0d139727cd39 100644 ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t)); if (inSeqs[idx].litLength) { -@@ -5642,26 +6370,15 @@ ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, +@@ -5644,26 +6372,15 @@ ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, ip += inSeqs[idx].litLength; seqPos->posInSrc += inSeqs[idx].litLength; } @@ -8536,7 +8536,7 @@ index f620cafca633..0d139727cd39 100644 { U32 idx = seqPos->idx; U32 startPosInSequence = seqPos->posInSequence; -@@ -5673,6 +6390,9 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5675,6 +6392,9 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* U32 bytesAdjustment = 0; U32 finalMatchSplit = 0; @@ -8546,7 +8546,7 @@ index f620cafca633..0d139727cd39 100644 if (cctx->cdict) { dictSize = cctx->cdict->dictContentSize; } else if (cctx->prefixDict.dict) { -@@ -5680,7 +6400,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5682,7 +6402,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* } else { dictSize = 0; } @@ -8555,7 +8555,7 @@ index f620cafca633..0d139727cd39 100644 DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) { -@@ -5688,7 +6408,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5690,7 +6410,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* U32 litLength = currSeq.litLength; U32 matchLength = currSeq.matchLength; U32 const rawOffset = currSeq.offset; @@ -8564,7 +8564,7 @@ index f620cafca633..0d139727cd39 100644 /* Modify the sequence depending on where endPosInSequence lies */ if (endPosInSequence >= currSeq.litLength + currSeq.matchLength) { -@@ -5702,7 +6422,6 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5704,7 +6424,6 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* /* Move to the next sequence */ endPosInSequence -= currSeq.litLength + currSeq.matchLength; startPosInSequence = 0; @@ -8572,7 +8572,7 @@ index f620cafca633..0d139727cd39 100644 } else { /* This is the final (partial) sequence we're adding from inSeqs, and endPosInSequence does not reach the end of the match. So, we have to split the sequence */ -@@ -5742,21 +6461,23 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5744,21 +6463,23 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* } /* Check if this offset can be represented with a repcode */ { U32 const ll0 = (litLength == 0); @@ -8603,7 +8603,7 @@ index f620cafca633..0d139727cd39 100644 } DEBUGLOG(5, "Ending seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); assert(idx == inSeqsSize || endPosInSequence <= inSeqs[idx].litLength + inSeqs[idx].matchLength); -@@ -5779,7 +6500,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* +@@ -5781,7 +6502,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* typedef size_t (*ZSTD_sequenceCopier) (ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos, const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, @@ -8612,7 +8612,7 @@ index f620cafca633..0d139727cd39 100644 static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode) { ZSTD_sequenceCopier sequenceCopier = NULL; -@@ -5793,6 +6514,57 @@ static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode) +@@ -5795,6 +6516,57 @@ static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode) return sequenceCopier; } @@ -8670,7 +8670,7 @@ index f620cafca633..0d139727cd39 100644 /* Compress, block-by-block, all of the sequences given. * * Returns the cumulative size of all compressed blocks (including their headers), -@@ -5805,9 +6577,6 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5807,9 +6579,6 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, const void* src, size_t srcSize) { size_t cSize = 0; @@ -8680,7 +8680,7 @@ index f620cafca633..0d139727cd39 100644 size_t remaining = srcSize; ZSTD_sequencePosition seqPos = {0, 0, 0}; -@@ -5827,22 +6596,29 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5829,22 +6598,29 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, } while (remaining) { @@ -8716,7 +8716,7 @@ index f620cafca633..0d139727cd39 100644 cSize += cBlockSize; ip += blockSize; op += cBlockSize; -@@ -5851,6 +6627,7 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5853,6 +6629,7 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, continue; } @@ -8724,7 +8724,7 @@ index f620cafca633..0d139727cd39 100644 compressedSeqsSize = ZSTD_entropyCompressSeqStore(&cctx->seqStore, &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy, &cctx->appliedParams, -@@ -5859,11 +6636,11 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5861,11 +6638,11 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, cctx->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, cctx->bmi2); FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed"); @@ -8738,7 +8738,7 @@ index f620cafca633..0d139727cd39 100644 /* We don't want to emit our first block as a RLE even if it qualifies because * doing so will cause the decoder (cli only) to throw a "should consume all input error." * This is only an issue for zstd <= v1.4.3 -@@ -5874,12 +6651,12 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5876,12 +6653,12 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, if (compressedSeqsSize == 0) { /* ZSTD_noCompressBlock writes the block header as well */ cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); @@ -8755,7 +8755,7 @@ index f620cafca633..0d139727cd39 100644 } else { U32 cBlockHeader; /* Error checking and repcodes update */ -@@ -5891,11 +6668,10 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5893,11 +6670,10 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, cBlockHeader = lastBlock + (((U32)bt_compressed)<<1) + (U32)(compressedSeqsSize << 3); MEM_writeLE24(op, cBlockHeader); cBlockSize = ZSTD_blockHeaderSize + compressedSeqsSize; @@ -8768,7 +8768,7 @@ index f620cafca633..0d139727cd39 100644 if (lastBlock) { break; -@@ -5906,12 +6682,15 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, +@@ -5908,12 +6684,15 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, dstCapacity -= cBlockSize; cctx->isFirstBlock = 0; } @@ -8785,7 +8785,7 @@ index f620cafca633..0d139727cd39 100644 const ZSTD_Sequence* inSeqs, size_t inSeqsSize, const void* src, size_t srcSize) { -@@ -5921,7 +6700,7 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapaci +@@ -5923,7 +6702,7 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapaci size_t frameHeaderSize = 0; /* Transparent initialization stage, same as compressStream2() */ @@ -8794,7 +8794,7 @@ index f620cafca633..0d139727cd39 100644 assert(cctx != NULL); FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed"); /* Begin writing output, starting with frame header */ -@@ -5949,26 +6728,34 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapaci +@@ -5951,26 +6730,34 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapaci cSize += 4; } @@ -8833,7 +8833,7 @@ index f620cafca633..0d139727cd39 100644 if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */ /* single thread mode : attempt to calculate remaining to flush more precisely */ { size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE; -@@ -6090,7 +6877,7 @@ static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, +@@ -6092,7 +6879,7 @@ static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, cp.targetLength = (unsigned)(-clampedCompressionLevel); } /* refine parameters based on srcSize & dictSize */ @@ -8842,7 +8842,7 @@ index f620cafca633..0d139727cd39 100644 } } -@@ -6125,3 +6912,29 @@ ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeH +@@ -6127,3 +6914,29 @@ ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeH if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN; return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown); } @@ -18615,7 +18615,7 @@ index 22686e367e6f..466828e35752 100644 MODULE_LICENSE("Dual BSD/GPL"); MODULE_DESCRIPTION("Zstd Common"); diff --git a/lib/zstd/zstd_compress_module.c b/lib/zstd/zstd_compress_module.c -index 04e1b5c01d9b..8ecf43226af2 100644 +index bd8784449b31..ceaf352d03e2 100644 --- a/lib/zstd/zstd_compress_module.c +++ b/lib/zstd/zstd_compress_module.c @@ -1,6 +1,6 @@ @@ -18627,7 +18627,7 @@ index 04e1b5c01d9b..8ecf43226af2 100644 * * This source code is licensed under both the BSD-style license (found in the diff --git a/lib/zstd/zstd_decompress_module.c b/lib/zstd/zstd_decompress_module.c -index f4ed952ed485..7d31518e9d5a 100644 +index 469fc3059be0..0ae819f0c927 100644 --- a/lib/zstd/zstd_decompress_module.c +++ b/lib/zstd/zstd_decompress_module.c @@ -1,6 +1,6 @@ @@ -18638,7 +18638,7 @@ index f4ed952ed485..7d31518e9d5a 100644 * All rights reserved. * * This source code is licensed under both the BSD-style license (found in the -@@ -77,7 +77,7 @@ EXPORT_SYMBOL(zstd_init_dstream); +@@ -113,7 +113,7 @@ EXPORT_SYMBOL(zstd_init_dstream); size_t zstd_reset_dstream(zstd_dstream *dstream) { @@ -18648,5 +18648,5 @@ index f4ed952ed485..7d31518e9d5a 100644 EXPORT_SYMBOL(zstd_reset_dstream); -- -2.44.0 +2.47.0.rc0 diff --git a/savedconfig/sys-kernel/linux-firmware-20240909-r1 b/savedconfig/sys-kernel/linux-firmware-20240909-r1 new file mode 100644 index 0000000..72b618c --- /dev/null +++ b/savedconfig/sys-kernel/linux-firmware-20240909-r1 @@ -0,0 +1,20 @@ +# Remove files that shall not be installed from this list. +amd-ucode/microcode_amd_fam19h.bin +amdgpu/navy_flounder_ce.bin +amdgpu/navy_flounder_dmcub.bin +amdgpu/navy_flounder_me.bin +amdgpu/navy_flounder_mec.bin +amdgpu/navy_flounder_mec2.bin +amdgpu/navy_flounder_pfp.bin +amdgpu/navy_flounder_rlc.bin +amdgpu/navy_flounder_sdma.bin +amdgpu/navy_flounder_smc.bin +amdgpu/navy_flounder_sos.bin +amdgpu/navy_flounder_ta.bin +amdgpu/navy_flounder_vcn.bin +v4l-cx231xx-avcore-01.fw +v4l-cx23418-apu.fw +v4l-cx23418-cpu.fw +v4l-cx23418-dig.fw +v4l-cx23885-avcore-01.fw +v4l-cx25840.fw diff --git a/sets/game-deps b/sets/game-deps index 0d31149..cd9ac4b 100644 --- a/sets/game-deps +++ b/sets/game-deps @@ -1,6 +1,10 @@ app-emulation/winetricks +app-emulation/protontricks + # winetricks-related stuff app-arch/cabextract app-arch/p7zip dev-libs/libappindicator # to fix systray icons from wine games -#gnome-extra/zenity + +# Some games need it (e.g. steamvr) +gnome-extra/zenity