On Mon, Nov 26, 2012 at 09:13:24PM -0800, Nitin Gupta wrote:

On 11/22/2012 06:42 PM, Minchan Kim wrote:

quoted

Lockdep complains about recursive deadlock of zram->init_lock.
Because zram_init_device could be called in reclaim context and
it requires a page with GFP_KERNEL.

We can fix it via replacing GFP_KERNEL with GFP_NOIO.
But more big problem is vzalloc in zram_init_device which calls GFP_KERNEL.
We can change it with __vmalloc which can receive gfp_t.
But still we have a problem. Although __vmalloc can handle gfp_t, it calls
allocation of GFP_KERNEL. That's why I sent the patch.
https://lkml.org/lkml/2012/4/23/77

Yes. Fundamental problem is utter crap API vmalloc.
If we can fix it, everyone would be happy. But life isn't simple
like seeing my thread of the patch.

So next option is to give up lazy initialization and initialize it at the
very disksize setting time. But it makes unnecessary metadata waste until
zram is really used. But let's think about it.

1) User of zram normally do mkfs.xxx or mkswap before using
   the zram block device(ex, normally, do it at booting time)
   It ends up allocating such metadata of zram before real usage so
   benefit of lazy initialzation would be mitigated.

2) Some user want to use zram when memory pressure is high.(ie, load zram
   dynamically, NOT booting time). It does make sense because people don't
   want to waste memory until memory pressure is high(ie, where zram is really
   helpful time). In this case, lazy initialzation could be failed easily
   because we will use GFP_NOIO instead of GFP_KERNEL for avoiding deadlock.
   So the benefit of lazy initialzation would be mitigated, too.

3) Metadata overhead is not critical and Nitin has a plan to diet it.
   4K : 12 byte(64bit machine) -> 64G : 192M so 0.3% isn't big overhead
   If insane user use such big zram device up to 20, it could consume 6% of ram
   but efficieny of zram will cover the waste.

So this patch gives up lazy initialization and instead we initialize metadata
at disksize setting time.

Signed-off-by: Minchan Kim <minchan@kernel.org>
---
 drivers/staging/zram/zram_drv.c   |   21 ++++-----------------
 drivers/staging/zram/zram_sysfs.c |    1 +
 2 files changed, 5 insertions(+), 17 deletions(-)

diff --git a/drivers/staging/zram/zram_drv.c b/drivers/staging/zram/zram_drv.c
index 9ef1eca..f364fb5 100644
--- a/drivers/staging/zram/zram_drv.c
+++ b/drivers/staging/zram/zram_drv.c

@@ -441,16 +441,13 @@ static void zram_make_request(struct request_queue *queue, struct bio *bio)
 {
 	struct zram *zram = queue->queuedata;

-	if (unlikely(!zram->init_done) && zram_init_device(zram))
-		goto error;
-
 	down_read(&zram->init_lock);
 	if (unlikely(!zram->init_done))
-		goto error_unlock;
+		goto error;

 	if (!valid_io_request(zram, bio)) {
 		zram_stat64_inc(zram, &zram->stats.invalid_io);
-		goto error_unlock;
+		goto error;
 	}

 	__zram_make_request(zram, bio, bio_data_dir(bio));

@@ -458,9 +455,8 @@ static void zram_make_request(struct request_queue *queue, struct bio *bio)

 	return;

-error_unlock:
-	up_read(&zram->init_lock);
 error:
+	up_read(&zram->init_lock);
 	bio_io_error(bio);
 }

@@ -509,19 +505,12 @@ void zram_reset_device(struct zram *zram)
 	up_write(&zram->init_lock);
 }

+/* zram->init_lock should be hold */

s/hold/held

Done.

btw, shouldn't we also change GFP_KERNEL to GFP_NOIO in
is_partial_io() case in both read/write handlers?

Absolutely. The previous patch isn't complete but sent by mistake.
Sorry for the noise.
I just sent new patch.

Thanks.

-- 
Kind regards,
Minchan Kim

--
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>