lkml.org 
[lkml]   [2005]   [Nov]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectWhat protection does sysfs_readdir have with SMP/Preemption?
From
Date
Hi,

I'm developing a custom kernel on top of Ingo's -rt patch. My kernel
makes race conditions in the vanilla kernel show up very well :-)

I just hit a bug, actually a page fault in fs/sysfs/dir.c in
sysfs_readdir:



for (p=q->next; p!= &parent_sd->s_children; p=p->next) {
struct sysfs_dirent *next;
const char * name;
int len;

next = list_entry(p, struct sysfs_dirent,
s_sibling);
if (!next->s_element)
continue;

name = sysfs_get_name(next);
len = strlen(name);
if (next->s_dentry)
ino = next->s_dentry->d_inode->i_ino;

^^^^
This is where I had a bad pointer reference.

else
ino = iunique(sysfs_sb, 2);

if (filldir(dirent, name, len, filp->f_pos, ino,
dt_type(next)) < 0)
return 0;


Looking at this code, I don't see anything protecting the s_dentry. For
example, couldn't the following happen:

sysfs_create_dir is called, which calls create_dir. Now we create a
dentry with no d_inode. In sysfs_make_dirent which calls
sysfs_new_dirent which adds to the parents s_children. Then
sysfs_make_dirent sets s_dentry = dentry (the one that was just made
with no d_inode assigned yet). Then create_dir calls sysfs_create which
finally assigns the d_inode.

So, either there is some hidden protection and my modification to the
kernel has caused this to bug, or we have just been lucky the whole time
in the vanilla kernel.

-- Steve


-
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: 2005-11-22 22:36    [W:0.060 / U:0.728 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site