[OpenAFS-devel] weird issue with reiserfs on 2.4.3 and openafs

[Neulinger, Nathan]

If I attempt to run afs on a afscache partition on reiserfs with 2.4.3 that
is already created (i.e. after the first time) I get this:

Apr 11 08:15:29 srvtst02 kernel: Starting AFS cache scan...Can't open inode
5
... oops snipped since it's just a trace to osi_panic in afsd...
Apr 11 08:15:29 srvtst02 kernel: 
Apr 11 08:15:36 srvtst02 kernel: vs-13048: reiserfs_iget: bad_inode. Stat
data of (2 5) not found
Apr 11 08:17:03 srvtst02 kernel: vs-13048: reiserfs_iget: bad_inode. Stat
data of (2 5) not found

The important part is that can't open inode 5, which corresponds to
/afscache/V0, which would be the first file afsd would attempt to open. 

Those vs-13048 errors come after afsd has segfaulted and I attempt to cat
/afscache/V0, which gives permission denied. Note - before I ran afsd, and
after I reboot, V0 is fine.

Note also - this ONLY happens after the first time. If I run it with the
initial empty /afscache, it works fine that first time only. Also works
fine. 

For testing I ran this with a 45MB cache, yielding 4500 cache nodes, which
should be well below any kernel inode limits, etc. 

This is the code that handles the open:

----------
void *osi_UFSOpen(ainode)
    afs_int32 ainode;
{
    struct inode *ip;
    register struct osi_file *afile = NULL;
    extern int cacheDiskType;
    afs_int32 code = 0;
    int dummy;
    struct inode *tip = NULL;
    struct file *filp = NULL;
    AFS_STATCNT(osi_UFSOpen);
    if(cacheDiskType != AFS_FCACHE_TYPE_UFS) {
    osi_Panic("UFSOpen called for non-UFS cache\n");
    }
    if (!afs_osicred_initialized) {
    /* valid for alpha_osf, SunOS, Ultrix */
    bzero((char *)&afs_osi_cred, sizeof(struct AFS_UCRED));
    crhold(&afs_osi_cred);  /* don't let it evaporate, since it is static */
    afs_osicred_initialized = 1;
    }
    afile = (struct osi_file *) osi_AllocSmallSpace(sizeof(struct
osi_file));
    AFS_GUNLOCK();
    if (!afile) {
    osi_Panic("osi_UFSOpen: Failed to allocate %d bytes for osi_file.\n",
           sizeof(struct osi_file));
    }
    memset(afile, 0, sizeof(struct osi_file));
    filp = &afile->file;
    filp->f_dentry = &afile->dentry;
    tip = iget(afs_cacheSBp, (u_long)ainode);
    if (!tip)
    osi_Panic("Can't get inode %d\n", ainode);
    FILE_INODE(filp) = tip;
    tip->i_flags |= MS_NOATIME; /* Disable updating access times. */
    filp->f_flags = O_RDWR;
#if defined(AFS_LINUX24_ENV)
    filp->f_op = fops_get(tip->i_fop);
#else
    filp->f_op = tip->i_op->default_file_ops;
#endif
    if (filp->f_op && filp->f_op->open)
    code = filp->f_op->open(tip, filp);
    if (code)
    osi_Panic("Can't open inode %d\n", ainode);
    afile->size = tip->i_size;
    AFS_GLOCK();
    afile->offset = 0;
    afile->proc = (int (*)()) 0;
    afile->inum = ainode;        /* for hint validity checking */
    return (void *)afile;
}
----------

So it seems to be that it's some issue with the f_op for open.

I modified the above code to print out code and it returns -5. Which appears
to correspond to EIO.

Any ideas?

-- Nathan

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-4841
Computing Services                       Fax: (573) 341-4216