On Fri, 3 Sep 2010 13:58:44 -0700
Andrew Morton <akpm at linux-foundation.org> wrote:
> On Thu, 2 Sep 2010 14:20:19 +0200
> Arnd Bergmann <arnd at arndb.de> wrote:
>
> > Commit 73296bc611 "procfs: Use generic_file_llseek in /proc/vmcore" broke
> > seeking on /proc/vmcore. This changes it back to use default_llseek
> > in order to restore the original behaviour.
>
> "broke" isn't a useful description of a bug :(
>
> A bit of list-trolling indicates that this change caused makedumpfile
> to hang during kexec dumping, yes? If so, why did this happen? How
> did the switch to generic_file_llseek() cause the /proc/vmcore seek
> behaviour to change?
>
> I'm not seeing a lot of difference between default_llseek() and
> generic_file_llseek() apart from file->f_mapping->host versus
> file->f_path.dentry->d_inode. Was it that, or what?
>
> etcetera. Better patch descriptions, please!
>
>
Ah...not related to this issue itself..but...
This is a patch for mmotm, which supports "unsigned file offset".
Now, mmotm, only /dev/mem support
==
From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
vfs-introduce-fmode_neg_offset-for-allowing-negative-f_pos.patch
adds "UNSIGINED" fpos support. But it dones't handle llseek.
Fix it.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu at jp.fujitsu.com>
---
fs/read_write.c | 33 ++++++++++++++++++---------------
1 file changed, 18 insertions(+), 15 deletions(-)
Index: mmotm-0827/fs/read_write.c
===================================================================
--- mmotm-0827.orig/fs/read_write.c
+++ mmotm-0827/fs/read_write.c
@@ -31,6 +31,20 @@ const struct file_operations generic_ro_
EXPORT_SYMBOL(generic_ro_fops);
+static int
+__negative_fpos_check(struct file *file, loff_t pos, size_t count)
+{
+ /*
+ * pos or pos+count is negative here, check overflow.
+ * too big "count" will be caught in rw_verify_area().
+ */
+ if ((pos < 0) && (pos + count < pos))
+ return -EOVERFLOW;
+ if (file->f_mode & FMODE_UNSIGNED_OFFSET)
+ return 0;
+ return -EINVAL;
+}
+
/**
* generic_file_llseek_unlocked - lockless generic llseek implementation
* @file: file structure to seek on
@@ -62,7 +76,9 @@ generic_file_llseek_unlocked(struct file
break;
}
- if (offset < 0 || offset > inode->i_sb->s_maxbytes)
+ if (offset < 0 && __negative_fpos_check(file, offset, 0))
+ return -EINVAL;
+ if (offset > inode->i_sb->s_maxbytes)
return -EINVAL;
/* Special lock needed here? */
@@ -137,7 +153,7 @@ loff_t default_llseek(struct file *file,
offset += file->f_pos;
}
retval = -EINVAL;
- if (offset >= 0) {
+ if (offset >= 0 || !__negative_fpos_check(file, offset, 0)) {
if (offset != file->f_pos) {
file->f_pos = offset;
file->f_version = 0;
@@ -222,19 +238,6 @@ bad:
}
#endif
-static int
-__negative_fpos_check(struct file *file, loff_t pos, size_t count)
-{
- /*
- * pos or pos+count is negative here, check overflow.
- * too big "count" will be caught in rw_verify_area().
- */
- if ((pos < 0) && (pos + count < pos))
- return -EOVERFLOW;
- if (file->f_mode & FMODE_UNSIGNED_OFFSET)
- return 0;
- return -EINVAL;
-}
/*
* rw_verify_area doesn't like huge counts. We limit
