mptcp: pm: Defer freeing of MPTCP userspace path manager entries
Subsystems (2)
Mat Martineau
maintainerPATCH v2 1/1 mptcp: pm: Defer freeing of MPTCP userspace path manager entries
When path manager entries are deleted from the local address list, they are first unlinked from the address list using list_del_rcu(). The entries must not be freed until after the RCU grace period, but the existing code immediately frees the entry.
Use kfree_rcu_mightsleep() and adjust sk_omem_alloc in open code instead of using the sock_kfree_s() helper. This code path is only called in a netlink handler, so the "might sleep" function is preferable to adding a rarely-used rcu_head member to struct mptcp_pm_addr_entry.
---
v2: Add comment explaining connection to sock_kfree_s, add "Fixes" tag
--- net/mptcp/pm_userspace.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
net/mptcp/pm_userspace.c +5 -1
@@ -337,7 +337,11 @@ int mptcp_pm_nl_remove_doit(struct sk_buff *skb, struct genl_info *info)release_sock (sk); -sock_kfree_s (sk, match,sizeof (*match)); +kfree_rcu_mightsleep (match); +/* Adjust sk_omem_alloc like sock_kfree_s() does, to match +* with allocation of this memory by sock_kmemdup() +*/ +atomic_sub (sizeof (*match), &sk->sk_omem_alloc); err =0 ; out:
Signature · 1 line
2.49.0
Hi Mat,
Thank you for your modifications, that's great!
Our CI did some validations and here is its report:
- KVM Validation: normal: Success! ✅
- KVM Validation: debug: Success! ✅
- KVM Validation: btf-normal (only bpftest_all): Success! ✅
- KVM Validation: btf-debug (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/14391915501
Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/b3df5ca41b93
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=952252
If there are some issues, you can reproduce them using the same environment as the one used by the CI thanks to a docker image, e.g.:
$ cd [kernel source code]
$ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
--pull always mptcp/mptcp-upstream-virtme-docker:latest \
auto-normalFor more details:
https://github.com/multipath-tcp/mptcp-upstream-virtme-docker
Please note that despite all the efforts that have been already done to have a stable tests suite when executed on a public CI like here, it is possible some reported issues are not due to your modifications. Still, do not hesitate to help us improve that ;-)
Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)
Hi Mat,
Quoted 10 lines · Mat Martineau
When path manager entries are deleted from the local address list, they are first unlinked from the address list using
list_del_rcu(). The entries must not be freed until after the RCU grace period, but the existing code immediately frees the entry.Use
kfree_rcu_mightsleep()and adjustsk_omem_allocin open code instead of using thesock_kfree_s()helper. This code path is only called in a netlink handler, so the "might sleep" function is preferable to adding a rarely-usedrcu_headmember to structmptcp_pm_addr_entry.
Quoted 5 lines
Fixes 88d097316371 ("mptcp: drop free_list for deleting entries")Signed-off-by Mat Martineau <[email protected]>---
v2: Add comment explaining connection to
sock_kfree_s, add "Fixes" tag
Thank you for the v2, and your reply on my previous review. This v2 looks good to me:
I guess I should wait to apply "only remove entry from local_addr_list when sending a REMOVE_ADDR" series first then, right? Are these other patches supposed to be fixes for net? They are not marked as that for the moment: in that case, should we add another patch to fix mptcp_userspace_pm_delete_local_addr() as well?
Cheers,
Matt
Signature · 1 line
Sponsored by the NGI0 Core fund.
Mat Martineau
maintainerQuoted 22 lines · Matthieu Baerts
Hi Mat,
Quoted 10 lines · Mat Martineau
When path manager entries are deleted from the local address list, they are first unlinked from the address list using
list_del_rcu(). The entries must not be freed until after the RCU grace period, but the existing code immediately frees the entry.Use
kfree_rcu_mightsleep()and adjustsk_omem_allocin open code instead of using thesock_kfree_s()helper. This code path is only called in a netlink handler, so the "might sleep" function is preferable to adding a rarely-usedrcu_headmember to structmptcp_pm_addr_entry.Quoted 5 lines
Fixes 88d097316371 ("mptcp: drop free_list for deleting entries")Signed-off-by Mat Martineau <[email protected]>---
v2: Add comment explaining connection to
sock_kfree_s, add "Fixes" tagThank you for the v2, and your reply on my previous review. This v2 looks good to me:
Quoted 8 lines
Reviewed-by Matthieu Baerts (NGI0) <[email protected]>I guess I should wait to apply "only remove entry from
local_addr_listwhen sending aREMOVE_ADDR" series first then, right? Are these other patches supposed to be fixes for net? They are not marked as that for the moment: in that case, should we add another patch to fixmptcp_userspace_pm_delete_local_addr()as well?
I don't think there's any issue or conflict with applying this to export-net first, it would leave the bug not fixed in mptcp_userspace_pm_delete_local_addr(). If this patch needs to stand alone, then it would need to be reworked to add a struct rcu_head to the local addr entry because mptcp_userspace_pm_delete_local_addr() gets called in contexts where sleepable calls aren't allowed.
I'll reply separately to v2 of the "only remove entry from local_addr_list when sending a REMOVE_ADDR" series.
- Mat
Matthieu Baerts
maintainerHi Mat,
Quoted 33 lines · Mat Martineau
Quoted 22 lines · Matthieu Baerts
Hi Mat,
Quoted 10 lines · Mat Martineau
When path manager entries are deleted from the local address list, they are first unlinked from the address list using
list_del_rcu(). The entries must not be freed until after the RCU grace period, but the existing code immediately frees the entry.Use
kfree_rcu_mightsleep()and adjustsk_omem_allocin open code instead of using thesock_kfree_s()helper. This code path is only called in a netlink handler, so the "might sleep" function is preferable to adding a rarely-usedrcu_headmember to structmptcp_pm_addr_entry.Quoted 5 lines
Fixes 88d097316371 ("mptcp: drop free_list for deleting entries")Signed-off-by Mat Martineau <[email protected]>---
v2: Add comment explaining connection to
sock_kfree_s, add "Fixes" tagThank you for the v2, and your reply on my previous review. This v2 looks good to me:
Quoted 8 lines
Reviewed-by Matthieu Baerts (NGI0) <[email protected]>I guess I should wait to apply "only remove entry from
local_addr_listwhen sending aREMOVE_ADDR" series first then, right? Are these other patches supposed to be fixes for net? They are not marked as that for the moment: in that case, should we add another patch to fixmptcp_userspace_pm_delete_local_addr()as well?
Quoted 6 lines
I don't think there's any issue or conflict with applying this to export-net first, it would leave the bug not fixed in
mptcp_userspace_pm_delete_local_addr(). If this patch needs to stand alone, then it would need to be reworked to add a structrcu_headto the local addr entry becausemptcp_userspace_pm_delete_local_addr()gets called in contexts where sleepable calls aren't allowed.
Thank you, it makes sense. Now in our tree:
New patches for t/upstream-net and t/upstream:
- 49746c77f0b1: mptcp: pm: Defer freeing of MPTCP userspace path manager entries
- Results: 0142ff39fa9b..ce1452067424 (export-net)
- Results: f1e32c11be27..309e5b3a6b32 (export)
Tests are now in progress:
- export-net:
https://github.com/multipath-tcp/mptcp_net-next/commit/cbf7c52183ef1d88ba762839fb579780ff6150e8/checks
- export:
https://github.com/multipath-tcp/mptcp_net-next/commit/b8b8b4ab98727371e1b3278b18314402cecd8da6/checks
Quoted 2 lines
I'll reply separately to v2 of the "only remove entry from
local_addr_listwhen sending aREMOVE_ADDR" series.
Thank you!
Cheers,
Matt
Signature · 1 line
Sponsored by the NGI0 Core fund.