[1/1] proc_keys_next should increase position index
diff mbox series

Message ID af9dcaa7-6e4f-281a-2bae-fb605cc55d2d@virtuozzo.com
State New
Headers show
Series
  • keys: seq_file .next functions should increase position index
Related show

Commit Message

Vasily Averin Jan. 24, 2020, 6:25 a.m. UTC
if seq_file .next fuction does not change position index,
read after some lseek can generate unexpected output.

https://bugzilla.kernel.org/show_bug.cgi?id=206283
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
---
 security/keys/proc.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

David Howells Jan. 27, 2020, 11:39 a.m. UTC | #1
I don't see the effect you're talking about with /proc/keys.  I see the
following:

	[root@andromeda ~]# dd if=/proc/keys bs=40 skip=1
	dd: /proc/keys: cannot skip to specified offset

and then it follows up with the normal content with no obvious duplicates (the
lines are numbered ascendingly in the first column).

I think I may be being confused by what you mean by "the last line".

David
Vasily Averin Jan. 27, 2020, 7:27 p.m. UTC | #2
On 1/27/20 2:39 PM, David Howells wrote:
> I don't see the effect you're talking about with /proc/keys.  I see the
> following:
> 
> 	[root@andromeda ~]# dd if=/proc/keys bs=40 skip=1
> 	dd: /proc/keys: cannot skip to specified offset
> 
> and then it follows up with the normal content with no obvious duplicates (the
> lines are numbered ascendingly in the first column).
> 
> I think I may be being confused by what you mean by "the last line".

on unpatched kernel

$ uname -a
Linux vvsx1 5.3.0-26-generic #28~18.04.1-Ubuntu SMP Wed Dec 18 16:40:14 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

$ dd if=/proc/keys bs=1  # VvS: full usual output
0f6bfdf5 I--Q---     2 perm 3f010000  1000  1000 user      4af2f79ab8848d0a: 740
1fb91b32 I--Q---     3 perm 1f3f0000  1000 65534 keyring   _uid.1000: 2
27589480 I--Q---     1 perm 0b0b0000     0     0 user      invocation_id: 16
2f33ab67 I--Q---   152 perm 3f030000     0     0 keyring   _ses: 2
33f1d8fa I--Q---     4 perm 3f030000  1000  1000 keyring   _ses: 1
3d427fda I--Q---     2 perm 3f010000  1000  1000 user      69ec44aec7678e5a: 740
3ead4096 I--Q---     1 perm 1f3f0000  1000 65534 keyring   _uid_ses.1000: 1
521+0 records in
521+0 records out
521 bytes copied, 0,00123769 s, 421 kB/s

$ dd if=/proc/keys bs=500 skip=1  # read after lseek in middle of last line
dd: /proc/keys: cannot skip to specified offset
g   _uid_ses.1000: 1        <<<< end of last line
3ead4096 I--Q---     1 perm 1f3f0000  1000 65534 keyring   _uid_ses.1000: 1   <<<< and whole last lien again
0+1 records in
0+1 records out
97 bytes copied, 0,000135035 s, 718 kB/s

$ dd if=/proc/keys bs=1000 skip=1   # read after lseek beyond end of file
dd: /proc/keys: cannot skip to specified offset
3ead4096 I--Q---     1 perm 1f3f0000  1000 65534 keyring   _uid_ses.1000: 1   <<<< generates last line
0+1 records in
0+1 records out
76 bytes copied, 0,000119981 s, 633 kB/s

On patched kernel:
[test@localhost ~]$ uname -a
Linux localhost.localdomain 5.5.0-rc6-00151-gd8d014f #8 SMP Fri Jan 24 13:25:06 MSK 2020 x86_64 x86_64 x86_64 GNU/Linux

[test@localhost ~]$ dd if=/proc/keys bs=1
06e8bec5 I--Q---     4 perm 1f3f0000  1000 65534 keyring   _uid.1000: empty
1b7ee8ed I--Q---    11 perm 3f030000  1000  1000 keyring   _ses: 1
2c1a365d I--Q---     1 perm 1f3f0000  1000 65534 keyring   _uid_ses.1000: 1
3f5823b4 I--Q---     6 perm 3f030000  1000  1000 keyring   _ses: 1
286+0 records in
286+0 records out
286 bytes copied, 0,000414581 s, 690 kB/s

[test@localhost ~]$ dd if=/proc/keys bs=270 skip=1  # VvS: read after lseek in middle of last line
dd: /proc/keys: cannot skip to specified offset
yring   _ses: 1   <<<< only end of last line was generated, as expected
0+1 records in
0+1 records out
16 bytes copied, 7,7199e-05 s, 207 kB/s

[test@localhost ~]$ dd if=/proc/keys bs=1000 skip=1   # VvS: read after lseek beond end of file
dd: /proc/keys: cannot skip to specified offset
0+0 records in   <<<< nothing was generated, as expected
0+0 records out
0 bytes copied, 8,8036e-05 s, 0,0 kB/s
Jarkko Sakkinen Jan. 30, 2020, 8:38 a.m. UTC | #3
On Fri, 2020-01-24 at 09:25 +0300, Vasily Averin wrote:
> if seq_file .next fuction does not change position index,
> read after some lseek can generate unexpected output.
> 
> https://bugzilla.kernel.org/show_bug.cgi?id=206283
> Signed-off-by: Vasily Averin <vvs@virtuozzo.com>

Similar comments as I gave here:

https://patchwork.kernel.org/patch/11346943/

/Jarkko
Jarkko Sakkinen Jan. 30, 2020, 8:42 a.m. UTC | #4
On Mon, 2020-01-27 at 11:39 +0000, David Howells wrote:
> I don't see the effect you're talking about with /proc/keys.  I see the
> following:
> 
> 	[root@andromeda ~]# dd if=/proc/keys bs=40 skip=1
> 	dd: /proc/keys: cannot skip to specified offset
> 
> and then it follows up with the normal content with no obvious duplicates (the
> lines are numbered ascendingly in the first column).
> 
> I think I may be being confused by what you mean by "the last line".
> 
> David


The commit message is completely lacking cause and effect. For
similar TPM commit I gave the following remarks:

https://patchwork.kernel.org/patch/11346943/

/Jarkko

Patch
diff mbox series

diff --git a/security/keys/proc.c b/security/keys/proc.c
index 415f3f1..d0cde66 100644
--- a/security/keys/proc.c
+++ b/security/keys/proc.c
@@ -139,6 +139,8 @@  static void *proc_keys_next(struct seq_file *p, void *v, loff_t *_pos)
 	n = key_serial_next(p, v);
 	if (n)
 		*_pos = key_node_serial(n);
+	else
+		(*_pos)++;
 	return n;
 }