Alex Romosan <[EMAIL PROTECTED]> writes:
> Russ Allbery <[EMAIL PROTECTED]> writes:
>> By vanilla, do you mean the stock Debian packages or building your own?
> by vanilla i mean the standard linus kernel from kernel.org that i
> built myself. i don't know how it would work for the standard debian
> kernel. that particular piece of openafs code is using tasklist_lock
> which is not available for non gpl modules any more in 2.6.18.
Sorry about the delay in investigating this further. I've been pretty
hammered lately with work.
I'm starting to suspect that there's something wrong with your
openafs-modules-source package or the untarring thereof that's causing you
to not get the latest source. I have no idea what that might be, but your
original bug report included the patch:
--- openafs/src/afs/afs_osi.c 2006-10-20 03:39:22.000000000 -0700
+++ openafs-keep/src/afs/afs_osi.c 2006-11-09 16:56:22.000000000
-0800
@@ -805,6 +805,7 @@
#endif
#if defined(AFS_LINUX22_ENV)
+extern rwlock_t tasklist_lock __attribute__((weak));
void
afs_osi_TraverseProcTable()
{
and my copy of openafs-modules-source has, around that line, the
following:
#if defined(AFS_LINUX22_ENV)
void
afs_osi_TraverseProcTable()
{
#if !defined(LINUX_KEYRING_SUPPORT)
extern rwlock_t tasklist_lock __attribute__((weak));
struct task_struct *p;
if (&tasklist_lock)
read_lock(&tasklist_lock);
In other words, that weak symbol is just four lines down from where you
added it, and so far as I can tell, all uses of tasklist_lock in that
routine are protected by #if !defined(LINUX_KEYRING_SUPPORT).
In src/afs/LINUX/osi_probe.c line 248 I have:
extern rwlock_t tasklist_lock __attribute__((weak));
before all uses of tasklist_lock in that file, and the uses of
tasklist_lock in osi_module.c all look to be ifdef'd out on Linux 2.4 or
later (Linux 2.4 and later, including 2.6, all define AFS_LINUX24_ENV).
Could you double-check that you're seeing the same source that I am? I
can't figure out how you're getting a reference to tasklist_lock without
the weak symbol definition and how your patch could have helped.
I just double-checked and the module builds and loads fine for me with the
Debian 2.6.18-2 kernel as well. I unfortunately don't have a system
that's running the stock upstream kernel to check.
--
Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
