[OpenAFS-devel] kernel BUG at /scratch/openafs/src/libafs/MODLOAD-2.6.13-MP/rx_kcommon.c:131!

[Martin MOKREJŠ]

phylo .phylo.natur.cuni.cz # /usr/vice/etc/afsd -shutdown -debug -verbose
afsd: Shutting down all afs processes and afs state
SScall(137, 28, 201)=13 afsd: AFS still mounted; Not shutting down
phylo .phylo.natur.cuni.cz # umount /afs
umount: /afs: device is busy
umount: /afs: device is busy
phylo .phylo.natur.cuni.cz # /usr/vice/etc/afsd -shutdown -debug -verbose
afsd: Shutting down all afs processes and afs state
SScall(137, 28, 201)=13 afsd: AFS still mounted; Not shutting down
phylo .phylo.natur.cuni.cz # cd /
phylo / # umount /afs
phylo / # dmesg
[cut]
Starting AFS cache scan...found 0 non-empty cache files (0%).
afs_dentry_iput("phylo.natur.cuni.cz", f6eaed64)
AFS isn't unmounted yet! Call aborted
AFS isn't unmounted yet! Call aborted
afs_dentry_iput("afsws", e7cf802c)
afs_dentry_iput("home", e7df2d64)
afs_dentry_iput("phylo.natur.cuni.cz", f6eae260)
afs_dentry_iput("usr", e7cf8d64)
afs_dentry_iput("i386_linux26", e7d12260)
afs_dentry_iput(".phylo.natur.cuni.cz", f6eaed64)
afs_dentry_iput("/", f6eae8fc)
COLD shutting down of: CB... afs... BkG... CTrunc... AFSDB... RxEvent... UnmaskRxkSignals... RxListener... 
phylo / # 


So, it looks like the right approach. Would you suggest some more tests
for the RW mounts? I often work with RW volumes.
M.



chas williams - CONTRACTOR wrote:
> In message <0177ADA1C8ECB841A21DD4AC@endicott>,Chaskiel M Grundman writes:
> 
>>I don't know. I don't really have a guess as to what the problem is. I'd
>>probably start by reintroducing the debugging code into
>>osi_linux_free_inode_pages. Perhaps also dump the old/new inode pointer
>>values and dp->d_count when the root changes so we know if it is the new or
>>old root that is the leftover inode.
> 
> 
> i did all that initially when i wrote the code.  but after looking at
> this with some fresh eyes i see that i forgot to dput() the reference
> i took from d_find_alias().  the following seems to work now without
> resetting d_count.
> 
> better?
> 
> Index: src/afs/afs_daemons.c
> ===================================================================
> RCS file: /cvs/openafs/src/afs/afs_daemons.c,v
> retrieving revision 1.33
> diff -u -u -r1.33 afs_daemons.c
> --- src/afs/afs_daemons.c	3 Apr 2005 18:13:30 -0000	1.33
> +++ src/afs/afs_daemons.c	7 Sep 2005 13:00:50 -0000
> @@ -309,8 +309,56 @@
>  		 * count to zero and fs checkv is executed when the current
>  		 * directory is /afs.
>  		 */
> +
> +#ifdef AFS_LINUX20_ENV
> +		{
> +			struct vrequest treq;
> +			struct vattr vattr;
> +			cred_t *credp;
> +			struct dentry *dp;
> +			struct vcache *vcp;
> +
> +			afs_rootFid.Fid.Volume = volid;
> +			afs_rootFid.Fid.Vnode = 1;
> +			afs_rootFid.Fid.Unique = 1;
> +
> +			credp = crref();
> +			if (afs_InitReq(&treq, credp))
> +			    goto out;
> +			vcp = afs_GetVCache(&afs_rootFid, &treq, NULL, NULL);
> +			if (!vcp)
> +			    goto out;
> +			afs_getattr(vcp, &vattr, credp);
> +			afs_fill_inode(AFSTOV(vcp), &vattr);
> +
> +			dp = d_find_alias(AFSTOV(afs_globalVp));
> +
> +#if defined(AFS_LINUX24_ENV)
> +			spin_lock(&dcache_lock);
> +#if defined(AFS_LINUX26_ENV)
> +			spin_lock(&dp->d_lock);
> +#endif
> +#endif
> +			list_del_init(&dp->d_alias);
> +			list_add(&dp->d_alias, &(AFSTOV(vcp)->i_dentry));
> +			dp->d_inode = AFSTOV(vcp);
> +#if defined(AFS_LINUX24_ENV)
> +#if defined(AFS_LINUX26_ENV)
> +			spin_unlock(&dp->d_lock);
> +#endif
> +			spin_unlock(&dcache_lock);
> +#endif
> +			dput(dp);
> +
> +			AFS_FAST_RELE(afs_globalVp);
> +			afs_globalVp = vcp;
> +out:
> +			crfree(credp);
> +		}
> +#else
>  		AFS_FAST_RELE(afs_globalVp);
>  		afs_globalVp = 0;
> +#endif
>  	    }
>  	    afs_rootFid.Fid.Volume = volid;
>  	    afs_rootFid.Fid.Vnode = 1;
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel
> 
> 

-- 
Martin Mokrejs
Email: 'bW9rcmVqc21Acmlib3NvbWUubmF0dXIuY3VuaS5jeg==\n'.decode('base64')
GPG key is at http://www.natur.cuni.cz/~mmokrejs