On Tue, Jan 22, 2013 at 10:47:17AM +0100, Jerome Marchand wrote:
> On 01/22/2013 01:07 AM, Minchan Kim wrote:
> > Now zram allocates new page with GFP_KERNEL in zram I/O path
> > if IO is partial. Unfortunately, It may cuase deadlock with
> > reclaim path so this patch solves the problem.
> >
> > Cc: Nitin Gupta <ngupta@xxxxxxxxxx>
> > Cc: Jerome Marchand <jmarchan@xxxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx
> > Signed-off-by: Minchan Kim <minchan@xxxxxxxxxx>
> > ---
> >
> > We could use GFP_IO instead of GFP_ATOMIC in zram_bvec_read with
> > some modification related to buffer allocation in case of partial IO.
> > But it needs more churn and prevent merge this patch into stable
> > if we should send this to stable so I'd like to keep it as simple
> > as possbile. GFP_IO usage could be separate patch after we merge it.
> > Thanks.
>
> I'd rather have a preallocated buffer for that. It would make
> zram_bvec_read/write() simpler (no need to deal with an allocation
> failure or to free the buffer) and it would be consistent with the way
> other similar buffer works (compress_workmem/buffer).
Consistent? Other buffers are MUST for zram working while the buffer
for partial I/O is supplement. Although partial I/O might be common in your config,
it doesn't match with my usecase. I didn't see any partial IO in my usecase.
Nontheless, why should I pay free 4K? Because of just making code SIMPLE?
I don't think current alloc/free handling about partial I/O is mess at the cost
of 4K. And we could use a few zram(a swap and a 2-compressed tmpfs) in system
so the cost is n*4K. Please keep in mind that ZRAM's goal is memory efficiency
and have used in many embedded system which they are always trying to save
just hundred byte.
>
> Jerome
>
> >
> > drivers/staging/zram/zram_drv.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/staging/zram/zram_drv.c b/drivers/staging/zram/zram_drv.c
> > index 61fb8f1..b285b3a 100644
> > --- a/drivers/staging/zram/zram_drv.c
> > +++ b/drivers/staging/zram/zram_drv.c
> > @@ -220,7 +220,7 @@ static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
> > user_mem = kmap_atomic(page);
> > if (is_partial_io(bvec))
> > /* Use a temporary buffer to decompress the page */
> > - uncmem = kmalloc(PAGE_SIZE, GFP_KERNEL);
> > + uncmem = kmalloc(PAGE_SIZE, GFP_ATOMIC);
> > else
> > uncmem = user_mem;
> >
> > @@ -268,7 +268,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
> > * This is a partial IO. We need to read the full page
> > * before to write the changes.
> > */
> > - uncmem = kmalloc(PAGE_SIZE, GFP_KERNEL);
> > + uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
> > if (!uncmem) {
> > pr_info("Error allocating temp memory!\n");
> > ret = -ENOMEM;
> >
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@xxxxxxxxx. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
--
Kind regards,
Minchan Kim
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>