Re: Trying to port data-logging to RH 2.4.18-19.7.x kernel

[John Dalbec <jpdalbec@ysu.edu>]

Chris Mason wrote:
> On Fri, 2003-01-31 at 16:28, John Dalbec wrote:
> 
> 
>>The immediate caller is the "ReiserFS specific hack" in 
>>fs/inode.c:get_inode signed <mason@suse.com>.  Is the BKL supposed to be 
>>held when get_inode is called?  
> 
> 
> Traditionally, the BKL is supposed to be held when iget or iget4 is
> called.  RedHat might have patches that do away with that and simply
> missed reiserfs, but it is more likely they have a patch to reduce BKL
> use in NFS that missed the iget4 case.

After inspecting nfsfh.c I think the "correct" fix is to grab the BKL in 
find_fh_dentry before calling nfsd_get_dentry.  There are other calls to 
nfsd_get_dentry later in find_fh_dentry but the BKL is already held at 
that point.

--- linux/fs/nfsd/nfsfh.c.orig  Tue Feb  4 10:55:36 2003
+++ linux/fs/nfsd/nfsfh.c       Tue Feb  4 10:58:19 2003
@@ -410,7 +410,9 @@
          */
   retry:
         down(&sb->s_nfsd_free_path_sem);
+       lock_kernel();
         result = nfsd_get_dentry(sb, datap, len, fhtype, 0);
+       unlock_kernel();
         if (IS_ERR(result)
             || !(result->d_flags & DCACHE_NFSD_DISCONNECTED)
             || (!S_ISDIR(result->d_inode->i_mode) && ! needpath)) {

> 
> So your two basic choices are adding the BKL to reiserfs_read_inode2, or
> going into the nfsd source and putting them around the iget4 call.  You
> might want to double check to see if their source had the BKL in
> reiserfs_read_inode2 before you started the data logging port.   
> 
> If not, you should be able to reproduce the oops on an unmodified redhat
> kernel (compiled with SMP on), and I'd appreciate it if you could send
> them a bug report as well.
> 
> -chris
> 
> 
>