sctp: revalidate list cursor after sctp_sendmsg_to_asoc() in SCTP_SENDALL
Subsystems (4)
PATCH 1/1 sctp: revalidate list cursor after sctp_sendmsg_to_asoc() in SCTP_SENDALL
From: Ben Morris <[email protected]>
The SCTP_SENDALL path in sctp_sendmsg() iterates ep->asocs with list_for_each_entry_safe(), which caches the next entry in @tmp before the loop body runs. The body calls sctp_sendmsg_to_asoc(), which may drop the socket lock inside sctp_wait_for_sndbuf().
While the lock is dropped, another thread can SCTP_SOCKOPT_PEELOFF the association cached in @tmp, migrating it to a new endpoint via sctp_sock_migrate() (list_del_init() + list_add_tail() to newep->asocs), and optionally close the new socket which frees the association via kfree_rcu(). The cached @tmp can also be freed by a network ABORT for that association, processed in softirq while the lock is dropped.
sctp_wait_for_sndbuf() revalidates @asoc (the current entry) on re-lock via the "sk != asoc->base.sk" and "asoc->base.dead" checks, but nothing revalidates @tmp. After a successful return, the iterator advances to the stale @tmp, yielding either a use-after-free (if the peeled socket was closed) or a list-walk onto the new endpoint's list head (type confusion of &newep->asocs as a struct sctp_association *).
Both are reachable from CapEff=0; the type-confusion path gives controlled indirect call via the outqueue.sched->init_sid pointer.
Fix by re-deriving @tmp from @asoc after sctp_sendmsg_to_asoc() returns. @asoc is known to still be on ep->asocs at that point: the only callers that list_del an association from ep->asocs are sctp_association_free() (which sets asoc->base.dead) and sctp_assoc_migrate() (which changes asoc->base.sk), and sctp_wait_for_sndbuf() checks both under the lock before any successful return; a tripped check propagates as err < 0 and the loop bails before the re-derive.
The SCTP_ABORT path in sctp_sendmsg_check_sflags() returns 0 and the loop hits 'continue' before sctp_sendmsg_to_asoc() is ever called, so the @tmp cached by list_for_each_entry_safe() still covers the lock-held free that ba59fb027307 ("sctp: walk the list of asoc safely") was added for.
--- net/sctp/socket.c | 9 +++++++++ 1 file changed, 9 insertions(+)
net/sctp/socket.c +9 -0
@@ -1986,6 +1986,15 @@ static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len) goto out_unlock;iov_iter_revert (&msg->msg_iter, err); + +/* sctp_sendmsg_to_asoc() may have released the socket +* lock (sctp_wait_for_sndbuf), during which other +* associations on ep->asocs could have been peeled +* off or freed. @asoc itself is revalidated by the +* base.dead and base.sk checks in sctp_wait_for_sndbuf, +* so re-derive the cached cursor from it. +*/ + tmp =list_next_entry (asoc, asocs); } goto out_unlock;
--
2.43.0
Quoted 41 lines · On Thu, May 7, 2026 at 8:15 PM <[email protected]> wrote:
From: Ben Morris <[email protected]>
The
SCTP_SENDALLpath insctp_sendmsg()iteratesep->asocswithlist_for_each_entry_safe(), which caches the next entry in @tmp before the loop body runs. The body callssctp_sendmsg_to_asoc(), which may drop the socket lock insidesctp_wait_for_sndbuf().While the lock is dropped, another thread can
SCTP_SOCKOPT_PEELOFFthe association cached in @tmp, migrating it to a new endpoint viasctp_sock_migrate()(list_del_init()+list_add_tail()tonewep->asocs), and optionally close the new socket which frees the association viakfree_rcu(). The cached @tmp can also be freed by a network ABORT for that association, processed in softirq while the lock is dropped.
sctp_wait_for_sndbuf()revalidates @asoc (the current entry) on re-lock via the "sk !=asoc->base.sk" and "asoc->base.dead" checks, but nothing revalidates @tmp. After a successful return, the iterator advances to the stale @tmp, yielding either a use-after-free (if the peeled socket was closed) or a list-walk onto the new endpoint's list head (type confusion of&newep->asocsas a structsctp_association*).Both are reachable from CapEff=0; the type-confusion path gives controlled indirect call via the
outqueue.sched->init_sidpointer.Fix by re-deriving @tmp from @asoc after
sctp_sendmsg_to_asoc()returns. @asoc is known to still be onep->asocsat that point: the only callers thatlist_delan association fromep->asocsaresctp_association_free()(which setsasoc->base.dead) andsctp_assoc_migrate()(which changesasoc->base.sk), andsctp_wait_for_sndbuf()checks both under the lock before any successful return; a tripped check propagates as err < 0 and the loop bails before the re-derive.The
SCTP_ABORTpath insctp_sendmsg_check_sflags()returns 0 and the loop hits 'continue' beforesctp_sendmsg_to_asoc()is ever called, so the @tmp cached bylist_for_each_entry_safe()still covers the lock-held free thatba59fb027307("sctp: walk the list of asoc safely") was added for.
Quoted 4 lines
Fixes 4910280503f3 ("sctp: add support for snd flag SCTP_SENDALL process in sendmsg")Assisted-by claude:mythosSigned-off-by Ben Morris <[email protected]>
Jakub Kicinski
maintainerQuoted 42 lines · On Fri, 8 May 2026 16:35:21 -0400 Xin Long wrote:
Quoted 41 lines · On Thu, May 7, 2026 at 8:15 PM <[email protected]> wrote:
From: Ben Morris <[email protected]>
The
SCTP_SENDALLpath insctp_sendmsg()iteratesep->asocswithlist_for_each_entry_safe(), which caches the next entry in @tmp before the loop body runs. The body callssctp_sendmsg_to_asoc(), which may drop the socket lock insidesctp_wait_for_sndbuf().While the lock is dropped, another thread can
SCTP_SOCKOPT_PEELOFFthe association cached in @tmp, migrating it to a new endpoint viasctp_sock_migrate()(list_del_init()+list_add_tail()tonewep->asocs), and optionally close the new socket which frees the association viakfree_rcu(). The cached @tmp can also be freed by a network ABORT for that association, processed in softirq while the lock is dropped.
sctp_wait_for_sndbuf()revalidates @asoc (the current entry) on re-lock via the "sk !=asoc->base.sk" and "asoc->base.dead" checks, but nothing revalidates @tmp. After a successful return, the iterator advances to the stale @tmp, yielding either a use-after-free (if the peeled socket was closed) or a list-walk onto the new endpoint's list head (type confusion of&newep->asocsas a structsctp_association*).Both are reachable from CapEff=0; the type-confusion path gives controlled indirect call via the
outqueue.sched->init_sidpointer.Fix by re-deriving @tmp from @asoc after
sctp_sendmsg_to_asoc()returns. @asoc is known to still be onep->asocsat that point: the only callers thatlist_delan association fromep->asocsaresctp_association_free()(which setsasoc->base.dead) andsctp_assoc_migrate()(which changesasoc->base.sk), andsctp_wait_for_sndbuf()checks both under the lock before any successful return; a tripped check propagates as err < 0 and the loop bails before the re-derive.The
SCTP_ABORTpath insctp_sendmsg_check_sflags()returns 0 and the loop hits 'continue' beforesctp_sendmsg_to_asoc()is ever called, so the @tmp cached bylist_for_each_entry_safe()still covers the lock-held free thatba59fb027307("sctp: walk the list of asoc safely") was added for.
Quoted 6 lines
Quoted 4 lines
Fixes 4910280503f3 ("sctp: add support for snd flag SCTP_SENDALL process in sendmsg")Assisted-by claude:mythosSigned-off-by Ben Morris <[email protected]>Acked-by Xin Long <[email protected]>
FWIW sashiko says there's more?
https://sashiko.dev/#/patchset/20260508001455.3137-1-joycathacker%40gmail.com
Xin Long
maintainerQuoted 49 lines · Jakub Kicinski
Quoted 47 lines · On Fri, 8 May 2026 16:35:21 -0400 Xin Long wrote:
Quoted 45 lines · On Thu, May 7, 2026 at 8:15 PM <[email protected]> wrote:
From: Ben Morris <[email protected]>
The
SCTP_SENDALLpath insctp_sendmsg()iteratesep->asocswithlist_for_each_entry_safe(), which caches the next entry in @tmp before the loop body runs. The body callssctp_sendmsg_to_asoc(), which may drop the socket lock insidesctp_wait_for_sndbuf().While the lock is dropped, another thread can
SCTP_SOCKOPT_PEELOFFthe association cached in @tmp, migrating it to a new endpoint viasctp_sock_migrate()(list_del_init()+list_add_tail()tonewep->asocs), and optionally close the new socket which frees the association viakfree_rcu(). The cached @tmp can also be freed by a network ABORT for that association, processed in softirq while the lock is dropped.
sctp_wait_for_sndbuf()revalidates @asoc (the current entry) on re-lock via the "sk !=asoc->base.sk" and "asoc->base.dead" checks, but nothing revalidates @tmp. After a successful return, the iterator advances to the stale @tmp, yielding either a use-after-free (if the peeled socket was closed) or a list-walk onto the new endpoint's list head (type confusion of&newep->asocsas a structsctp_association*).Both are reachable from CapEff=0; the type-confusion path gives controlled indirect call via the
outqueue.sched->init_sidpointer.Fix by re-deriving @tmp from @asoc after
sctp_sendmsg_to_asoc()returns. @asoc is known to still be onep->asocsat that point: the only callers thatlist_delan association fromep->asocsaresctp_association_free()(which setsasoc->base.dead) andsctp_assoc_migrate()(which changesasoc->base.sk), andsctp_wait_for_sndbuf()checks both under the lock before any successful return; a tripped check propagates as err < 0 and the loop bails before the re-derive.The
SCTP_ABORTpath insctp_sendmsg_check_sflags()returns 0 and the loop hits 'continue' beforesctp_sendmsg_to_asoc()is ever called, so the @tmp cached bylist_for_each_entry_safe()still covers the lock-held free thatba59fb027307("sctp: walk the list of asoc safely") was added for.Fixes 4910280503f3 ("sctp: add support for snd flag SCTP_SENDALL process in sendmsg")Assisted-by claude:mythosSigned-off-by Ben Morris <[email protected]>
Quoted 5 lines
Quoted 1 line
Acked-by Xin Long <[email protected]>FWIW sashiko says there's more?
https://sashiko.dev/#/patchset/20260508001455.3137-1-joycathacker%40gmail.com
I will try to verify this and submit a fix if it could be reproduced.
Thanks.
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:
Quoted 8 lines · On Thu, 7 May 2026 17:14:55 -0700 you wrote:
From: Ben Morris <[email protected]>
The
SCTP_SENDALLpath insctp_sendmsg()iteratesep->asocswithlist_for_each_entry_safe(), which caches the next entry in @tmp before the loop body runs. The body callssctp_sendmsg_to_asoc(), which may drop the socket lock insidesctp_wait_for_sndbuf().[...]
Here is the summary with links:
- [net] sctp: revalidate list cursor after sctp_sendmsg_to_asoc() in SCTP_SENDALL
https://git.kernel.org/netdev/net/c/abb5f36771cc
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html