Page MenuHomeFreeBSD

urtwn(4): fix mbuf leak in the TX path
ClosedPublic

Authored by avos on Oct 22 2015, 9:52 PM.
Tags
None
Referenced Files
F106383742: D3988.diff
Sun, Dec 29, 7:08 PM
Unknown Object (File)
Nov 26 2024, 11:38 AM
Unknown Object (File)
Nov 20 2024, 10:23 AM
Unknown Object (File)
Oct 28 2024, 9:48 PM
Unknown Object (File)
Oct 28 2024, 9:48 PM
Unknown Object (File)
Oct 28 2024, 9:48 PM
Unknown Object (File)
Oct 28 2024, 9:48 PM
Unknown Object (File)
Oct 28 2024, 9:32 PM
Subscribers

Details

Test Plan

Tested with RTL8188EU, STA mode

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

avos retitled this revision from to urtwn(4): fix mbuf leak in the TX path.
avos updated this object.
avos edited the test plan for this revision. (Show Details)
avos added reviewers: adrian, kevlo.
avos set the repository for this revision to rS FreeBSD src repository - subversion.

Hi,

I'd rather than we build a mostly sensible, consistent way of dealing with things with the USB drivers. Here, I'd like txeof to take an mbuf/txbuf and free it. That way all of the dequeue logic is done in the USB tx callback and it's not broken out. When we start seeing drivers do both USB and non-USB then i'd like there to be a shared txeof that makes no actual assumption that the underlying queuing is done via USB.

However yeah, the USB completion callback can leak buffers in specific instances. So let's fix that and keep XXX_txeof() take the softc and the txbuf to complete.

avos edited edge metadata.

Convert urtwn_txeof() to the common appearance.

kevlo edited edge metadata.
This revision is now accepted and ready to land.Oct 24 2015, 2:27 PM
adrian edited edge metadata.
This revision was automatically updated to reflect the committed changes.