lkml.org 
[lkml]   [2003]   [Feb]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectPossible bug in ext3 versus filter drivers in 2.4.18-3, 2.4.18-14 and 2.4.20.
Date
From
Hi list.

I am working on a filter driver.

In my driver, I am monitoring whether requests I redirected to other driver were successful or not. To do so, I am replacing b_end_io and b_private fields in buffer header. This way, instead of calling the default completion routine, system is calling my completion routine, which used to, among the other things, recover original b_end_io and b_private fields from value I placed in b_private when mapped the request and call original b_end_io with appropriate uptodate value.

The problem with ext3 is that it is accessing b_private field in locked buffer headers. It is treating b_private field I placed in buffer header, as journal header. As a result, I am getting multiple segmentation faults in different places and you can imagine what else.

The problem starts somewhere in ext3_new_block() in fs/ext3/balloc.c. In the begging it's obtaining buffer header and eventually it's calling __ext3_journal_get_undo_access() in include/linux/ext3_jbd.h. From there, it goes to journal_get_undo_access() in fs/jbd/transaction.c, then to journal_add_journal_head() in fs/jbd/journal.c and so on. Journal header is obtained in line stating "jh = bh2jh(bh);" in journal_add_journal_head().

I see two possible fixes. First, we can make sure filter drivers do not change b_private field in buffer header. It seems to be quite odd solution since as far as I understood, this is what b_private filed is there for (among the other things of course). Other option is to make sure that ext3_new_block() won't access locked buffer headers. This seems to be more reasonable.
I am afraid I am not really an expert in file systems in general and in ext3 in particularly, so I don't know what exactly to do. Perhaps someone can fix this thing or guide me how to do so.

Finally, I found this thing in 2.4.18-3 (RH 7.3). I checked 2.4.18-14 (RH 8.0) and 2.4.20. It seems that the problem is there for all three versions of kernel.

Thank you.

Alexandr Sandler.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:33    [W:0.033 / U:3.848 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site