[Bug 237921] wpi: Memory leak in function wpi_free_tx_ring of sys/dev/wpi/if_wpi.c

From: <bugzilla-noreply_at_freebsd.org>
Date: Sat, 11 May 2024 19:48:41 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=237921

Bjoern A. Zeeb <bz@FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bz@FreeBSD.org
             Status|Open                        |In Progress

--- Comment #9 from Bjoern A. Zeeb <bz@FreeBSD.org> ---
(In reply to Yusuf Khan from comment #8)

I cannot follow the logic.

wpi_free_tx_ring() is called from device detach (ignoring the attach error
path);  the mbuf/ring descriptor will not be used beyond that function anymore
at that stage so there is no need to set it to NULL anymore.  That would just
be a NOP.

If there's still a reference to a ni at that stage, then other cleanup hasn't
worked at the beginning of wpi_detach().  There should be no more ni;  in fact
that should also be no more mbuf attached on the TX ring so I would even
question if that code path is ever taken.

This function is basically doing the opposite of wpi_alloc_tx_ring().

wpi_reset_tx_ring() on the contrary leaves all resources in place and only
cleans up any outstanding transations and resets the ring to default values; 
this happens in wpi_hw_stop().  And then the resources will be re-used again
after wpi_parent() goes through the bringup part again (a VAP/wlan interface is
brought up again).


So let me ask what lead to the original assumption that there may be a memory
leak and a possible use-after-free.

If you indeed believe that, I would love to know the code path which can
trigger it.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are on the CC list for the bug.