On Thu, Aug 17, 2023 at 09:33:19PM +0200, Oliver Hartkopp wrote:
On 17.08.23 20:24, Jakub Kicinski wrote:
On Thu, 17 Aug 2023 10:58:36 -0400 Sasha Levin wrote:
This is a note to let you know that I've just added the patch titled
can: raw: fix receiver memory leak
I may be wrong but I _think_ here's still an unresolved deadlock
introduced by this change, even after Eric's follow up?
Yes, you are right!
The original patch
"can: raw: fix receiver memory leak"
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/net/can/raw.c?id=ee8b94c8510ce64afe0b87ef548d23e00915fb10
1. introduced a lockdep issue which was fixed here:
"can: raw: fix lockdep issue in raw_release()"
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/net/can/raw.c?id=11c9027c983e9e4b408ee5613b6504d24ebd85be
2. introduced a device refcount issue found by Eric Dumazet here:
https://lore.kernel.org/linux-can/CANn89iLXk-=Zh9va4eKM+35vx4=sn8p_Jvu5xAkHphOyh07B7w@xxxxxxxxxxxxxx/
I provided a fix for issue 2 here:
https://lore.kernel.org/linux-can/20230815171447.10879-1-socketcan@xxxxxxxxxxxx/T/#u
which needs to be reviewed and upstreamed now.
Until the two fixes for the original
"can: raw: fix receiver memory leak" patch are not in Linus' tree, we
should not add the patches at all.
Ack, I missed the second issue, will drop for now.
--
Thanks,
Sasha
