undefined shift in wb_update_dirty_ratelimit()

undefined shift in wb_update_dirty_ratelimit()

[Andrey Ryabinin]

I've hit undefined shift in wb_update_dirty_ratelimit() which does some
mysterious 'step' calculations:

	/*
	 * Don't pursue 100% rate matching. It's impossible since the balanced
	 * rate itself is constantly fluctuating. So decrease the track speed
	 * when it gets close to the target. Helps eliminate pointless tremors.
	 */
	step >>= dirty_ratelimit / (2 * step + 1);


dirty_ratelimit = INIT_BW and step = 0 results in this:

[ 5006.957366] ================================================================================
[ 5006.957798] UBSAN: Undefined behaviour in ../mm/page-writeback.c:1286:7
[ 5006.958091] shift exponent 25600 is too large for 64-bit type 'long unsigned int'
[ 5006.958414] CPU: 2 PID: 7452 Comm: trinity-c2 Not tainted 4.4.0-rc1+ #19
[ 5006.958740] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.8.2-0-g33fbe13 by qemu-project.org 04/01/2014
[ 5006.959247]  ffffffff8261427a ffff88007ddaf798 ffffffff815e6cd6 0000000000000001
[ 5006.959630]  ffff88007ddaf7c8 ffff88007ddaf7b0 ffffffff8163a62d ffffffff8261427a
[ 5006.959997]  ffff88007ddaf850 ffffffff8163accf ffff88013afac780 0000000000000202
[ 5006.960345] Call Trace:
[ 5006.960460]  [<ffffffff815e6cd6>] dump_stack+0x45/0x5f
[ 5006.960723]  [<ffffffff8163a62d>] ubsan_epilogue+0xd/0x40
[ 5006.960961]  [<ffffffff8163accf>] __ubsan_handle_shift_out_of_bounds+0xef/0x130
[ 5006.961282]  [<ffffffff8140555c>] ? __es_insert_extent+0x2ec/0x670
[ 5006.961554]  [<ffffffff815f19f3>] ? radix_tree_lookup_slot+0x13/0x30
[ 5006.961867]  [<ffffffff81218431>] __wb_update_bandwidth.constprop.26+0x521/0x6a0
[ 5006.962190]  [<ffffffff81219c67>] balance_dirty_pages.isra.23+0xa27/0x1900
[ 5006.962498]  [<ffffffff814174b7>] ? jbd2_journal_stop+0x237/0x6b0
[ 5006.962835]  [<ffffffff81301d1f>] ? __block_commit_write.isra.23+0x6f/0x140
[ 5006.963140]  [<ffffffff8121aca3>] balance_dirty_pages_ratelimited+0x163/0x340
[ 5006.963453]  [<ffffffff81204954>] generic_perform_write+0x184/0x290
[ 5006.963750]  [<ffffffff81206bb7>] __generic_file_write_iter+0x1b7/0x3b0
[ 5006.964064]  [<ffffffff8137fdc4>] ext4_file_write_iter+0x154/0x930
[ 5006.964336]  [<ffffffff812ad414>] vfs_iter_write+0xa4/0x140
[ 5006.964581]  [<ffffffff812fa32a>] iter_file_splice_write+0x27a/0x4f0
[ 5006.964881]  [<ffffffff812f8c13>] direct_splice_actor+0x53/0xd0
[ 5006.965141]  [<ffffffff812f91f6>] splice_direct_to_actor+0xf6/0x390
[ 5006.965417]  [<ffffffff81526c11>] ? security_file_permission+0x41/0x110
[ 5006.965708]  [<ffffffff812f8bc0>] ? wakeup_pipe_writers+0x60/0x60
[ 5006.965976]  [<ffffffff812f9529>] do_splice_direct+0x99/0x100
[ 5006.966228]  [<ffffffff81526c11>] ? security_file_permission+0x41/0x110
[ 5006.966539]  [<ffffffff812af36b>] do_sendfile+0x18b/0x6a0
[ 5006.966825]  [<ffffffff812b0431>] compat_SyS_sendfile+0x71/0x80
[ 5006.967092]  [<ffffffff81004875>] do_syscall_32_irqs_off+0x75/0x110
[ 5006.967382]  [<ffffffff81e35903>] entry_INT80_compat+0x33/0x40
[ 5006.967653] ================================================================================

Re: undefined shift in wb_update_dirty_ratelimit()

[Tejun Heo]

Hello, Andrey.

On Mon, Dec 07, 2015 at 05:17:06PM +0300, Andrey Ryabinin wrote:
> I've hit undefined shift in wb_update_dirty_ratelimit() which does some
> mysterious 'step' calculations:
> 
> 	/*
> 	 * Don't pursue 100% rate matching. It's impossible since the balanced
> 	 * rate itself is constantly fluctuating. So decrease the track speed
> 	 * when it gets close to the target. Helps eliminate pointless tremors.
> 	 */
> 	step >>= dirty_ratelimit / (2 * step + 1);
> 
> 
> dirty_ratelimit = INIT_BW and step = 0 results in this:
> 
> [ 5006.957366] ================================================================================
> [ 5006.957798] UBSAN: Undefined behaviour in ../mm/page-writeback.c:1286:7
> [ 5006.958091] shift exponent 25600 is too large for 64-bit type 'long unsigned int'

We prolly should do sth like

	shift = dirty_ratelimit / (2 * step = 1);
	if (shift < BITS_PER_LONG) {
		step = (step >> shift) + 7 / 8;
	} else {
		step = 0;
	}

Thanks.

-- 
tejun

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: undefined shift in wb_update_dirty_ratelimit()

[Andy Shevchenko]

On Mon, Dec 7, 2015 at 4:26 PM, Tejun Heo <tj@kernel.org> wrote:
> Hello, Andrey.
> On Mon, Dec 07, 2015 at 05:17:06PM +0300, Andrey Ryabinin wrote:
>> I've hit undefined shift in wb_update_dirty_ratelimit() which does some
>> mysterious 'step' calculations:
>>
>>       /*
>>        * Don't pursue 100% rate matching. It's impossible since the balanced
>>        * rate itself is constantly fluctuating. So decrease the track speed
>>        * when it gets close to the target. Helps eliminate pointless tremors.
>>        */
>>       step >>= dirty_ratelimit / (2 * step + 1);
>>
>>
>> dirty_ratelimit = INIT_BW and step = 0 results in this:
>>
>> [ 5006.957366] ================================================================================
>> [ 5006.957798] UBSAN: Undefined behaviour in ../mm/page-writeback.c:1286:7
>> [ 5006.958091] shift exponent 25600 is too large for 64-bit type 'long unsigned int'
>
> We prolly should do sth like
>
>         shift = dirty_ratelimit / (2 * step = 1);

>         if (shift < BITS_PER_LONG) {
>                 step = (step >> shift) + 7 / 8;

7/8 = 0. Missed parens?
In that case it's DIV_ROUND_UP(step >> shift, 8)


>         } else {
>                 step = 0;
>         }


-- 
With Best Regards,
Andy Shevchenko

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH] mm/page-writeback: fix dirty_ratelimit calculation

[Andrey Ryabinin]

Calculation of dirty_ratelimit sometimes is not correct.
E.g. initial values of dirty_ratelimit == INIT_BW and step == 0,
lead to the following result:

   UBSAN: Undefined behaviour in ../mm/page-writeback.c:1286:7
   shift exponent 25600 is too large for 64-bit type 'long unsigned int'

The fix is straightforward - make step 0 if the shift exponent is too big.

Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
---
 mm/page-writeback.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 6fe7d15..d782cba 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -1169,6 +1169,7 @@ static void wb_update_dirty_ratelimit(struct dirty_throttle_control *dtc,
 	unsigned long balanced_dirty_ratelimit;
 	unsigned long step;
 	unsigned long x;
+	unsigned long shift;
 
 	/*
 	 * The dirty rate will match the writeout rate in long term, except
@@ -1293,11 +1294,11 @@ static void wb_update_dirty_ratelimit(struct dirty_throttle_control *dtc,
 	 * rate itself is constantly fluctuating. So decrease the track speed
 	 * when it gets close to the target. Helps eliminate pointless tremors.
 	 */
-	step >>= dirty_ratelimit / (2 * step + 1);
-	/*
-	 * Limit the tracking speed to avoid overshooting.
-	 */
-	step = (step + 7) / 8;
+	shift = dirty_ratelimit / (2 * step + 1);
+	if (shift < BITS_PER_LONG)
+		step = DIV_ROUND_UP(step >> shift, 8);
+	else
+		step = 0;
 
 	if (dirty_ratelimit < balanced_dirty_ratelimit)
 		dirty_ratelimit += step;
-- 
2.4.10

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>