| commit | 9b0b4d8ae82cadd059d01ab93babcaa5e2054b93 | |
| author | Michael Buesch <mb@bu3sch.de> | |
| Thu, 6 Apr 2006 23:42:55 +0000 (7 01:42 +0200) | ||
| committer | John W. Linville <linville@tuxdriver.com> | |
| Wed, 19 Apr 2006 21:25:38 +0000 (19 17:25 -0400) | ||
| tree | 065d53b0c271a49660e356e39c94fbda218e1149 | treesnapshot (tar.gz zip) |
| parent | c1783454a31e05b94774951b0b5d1eb9075ebfb4 | commitdiff |
[PATCH] softmac: fix spinlock recursion on reassoc
This fixes a spinlock recursion on receiving a reassoc request.
On reassoc, the softmac calls back into the driver. This results in a
driver lock recursion. This schedules the assoc workqueue, instead
of calling it directly.
Probably, we should defer the _whole_ management frame processing
to a tasklet or workqueue, because it does several callbacks into the driver.
That is dangerous.
This fix should go into linus's tree, before 2.6.17 is released, because it
is remote exploitable (DoS by crash).
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This fixes a spinlock recursion on receiving a reassoc request.
On reassoc, the softmac calls back into the driver. This results in a
driver lock recursion. This schedules the assoc workqueue, instead
of calling it directly.
Probably, we should defer the _whole_ management frame processing
to a tasklet or workqueue, because it does several callbacks into the driver.
That is dangerous.
This fix should go into linus's tree, before 2.6.17 is released, because it
is remote exploitable (DoS by crash).
Signed-off-by: John W. Linville <linville@tuxdriver.com>
| net/ieee80211/softmac/ieee80211softmac_assoc.c | diffblobblamehistory |