From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- Documentation/networking/tproxy.rst | 109 ++++++++++++++++++++++++++++++++++++ 1 file changed, 109 insertions(+) create mode 100644 Documentation/networking/tproxy.rst (limited to 'Documentation/networking/tproxy.rst') diff --git a/Documentation/networking/tproxy.rst b/Documentation/networking/tproxy.rst new file mode 100644 index 000000000..75e4990cc --- /dev/null +++ b/Documentation/networking/tproxy.rst @@ -0,0 +1,109 @@ +.. SPDX-License-Identifier: GPL-2.0 + +========================= +Transparent proxy support +========================= + +This feature adds Linux 2.2-like transparent proxy support to current kernels. +To use it, enable the socket match and the TPROXY target in your kernel config. +You will need policy routing too, so be sure to enable that as well. + +From Linux 4.18 transparent proxy support is also available in nf_tables. + +1. Making non-local sockets work +================================ + +The idea is that you identify packets with destination address matching a local +socket on your box, set the packet mark to a certain value:: + + # iptables -t mangle -N DIVERT + # iptables -t mangle -A PREROUTING -p tcp -m socket --transparent -j DIVERT + # iptables -t mangle -A DIVERT -j MARK --set-mark 1 + # iptables -t mangle -A DIVERT -j ACCEPT + +Alternatively you can do this in nft with the following commands:: + + # nft add table filter + # nft add chain filter divert "{ type filter hook prerouting priority -150; }" + # nft add rule filter divert meta l4proto tcp socket transparent 1 meta mark set 1 accept + +And then match on that value using policy routing to have those packets +delivered locally:: + + # ip rule add fwmark 1 lookup 100 + # ip route add local 0.0.0.0/0 dev lo table 100 + +Because of certain restrictions in the IPv4 routing output code you'll have to +modify your application to allow it to send datagrams _from_ non-local IP +addresses. All you have to do is enable the (SOL_IP, IP_TRANSPARENT) socket +option before calling bind:: + + fd = socket(AF_INET, SOCK_STREAM, 0); + /* - 8< -*/ + int value = 1; + setsockopt(fd, SOL_IP, IP_TRANSPARENT, &value, sizeof(value)); + /* - 8< -*/ + name.sin_family = AF_INET; + name.sin_port = htons(0xCAFE); + name.sin_addr.s_addr = htonl(0xDEADBEEF); + bind(fd, &name, sizeof(name)); + +A trivial patch for netcat is available here: +http://people.netfilter.org/hidden/tproxy/netcat-ip_transparent-support.patch + + +2. Redirecting traffic +====================== + +Transparent proxying often involves "intercepting" traffic on a router. This is +usually done with the iptables REDIRECT target; however, there are serious +limitations of that method. One of the major issues is that it actually +modifies the packets to change the destination address -- which might not be +acceptable in certain situations. (Think of proxying UDP for example: you won't +be able to find out the original destination address. Even in case of TCP +getting the original destination address is racy.) + +The 'TPROXY' target provides similar functionality without relying on NAT. Simply +add rules like this to the iptables ruleset above:: + + # iptables -t mangle -A PREROUTING -p tcp --dport 80 -j TPROXY \ + --tproxy-mark 0x1/0x1 --on-port 50080 + +Or the following rule to nft:: + + # nft add rule filter divert tcp dport 80 tproxy to :50080 meta mark set 1 accept + +Note that for this to work you'll have to modify the proxy to enable (SOL_IP, +IP_TRANSPARENT) for the listening socket. + +As an example implementation, tcprdr is available here: +https://git.breakpoint.cc/cgit/fw/tcprdr.git/ +This tool is written by Florian Westphal and it was used for testing during the +nf_tables implementation. + +3. Iptables and nf_tables extensions +==================================== + +To use tproxy you'll need to have the following modules compiled for iptables: + + - NETFILTER_XT_MATCH_SOCKET + - NETFILTER_XT_TARGET_TPROXY + +Or the floowing modules for nf_tables: + + - NFT_SOCKET + - NFT_TPROXY + +4. Application support +====================== + +4.1. Squid +---------- + +Squid 3.HEAD has support built-in. To use it, pass +'--enable-linux-netfilter' to configure and set the 'tproxy' option on +the HTTP listener you redirect traffic to with the TPROXY iptables +target. + +For more information please consult the following page on the Squid +wiki: http://wiki.squid-cache.org/Features/Tproxy4 -- cgit v1.3.1