Xen devel - Jun 2013 - [PATCH 3/4] xen-blkback: check the number of iovecs before allocating a bios

With the introduction of indirect segments we can receive requests
with a number of segments bigger than the maximum number of allowed
iovecs in a bios, so make sure that blkback doesn't try to allocate a
bios with more iovecs than BIO_MAX_PAGES

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
---
 drivers/block/xen-blkback/blkback.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/drivers/block/xen-blkback/blkback.c
b/drivers/block/xen-blkback/blkback.c
index d622d86..876116b 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -1236,7 +1236,8 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
 				     seg[i].nsec << 9,
 				     seg[i].offset) == 0)) {
 
-			bio = bio_alloc(GFP_KERNEL, nseg-i);
+			int nr_iovecs = (nseg-i) > BIO_MAX_PAGES ? BIO_MAX_PAGES : (nseg-i);
+			bio = bio_alloc(GFP_KERNEL, nr_iovecs);
 			if (unlikely(bio == NULL))
 				goto fail_put_bio;
 
-- 
1.7.7.5 (Apple Git-26)


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

>>> On 21.06.13 at 12:56, Roger Pau Monne <roger.pau@citrix.com>
wrote:
> @@ -1236,7 +1236,8 @@ static int dispatch_rw_block_io(struct xen_blkif
*blkif,
>  				     seg[i].nsec << 9,
>  				     seg[i].offset) == 0)) {
>  
> -			bio = bio_alloc(GFP_KERNEL, nseg-i);
> +			int nr_iovecs = (nseg-i) > BIO_MAX_PAGES ? BIO_MAX_PAGES : (nseg-i);
I''m sure this results in a compiler warning (declaration after
statement).

And it surely would read much better if you used some form of
min() - the shorter line would at once allow you to properly
blank separate operands from operators.

Jan
> +			bio = bio_alloc(GFP_KERNEL, nr_iovecs);
>  			if (unlikely(bio == NULL))
>  				goto fail_put_bio;
>

On 21/06/13 13:46, Jan Beulich wrote:
>>>> On 21.06.13 at 12:56, Roger Pau Monne
<roger.pau@citrix.com> wrote:
>> @@ -1236,7 +1236,8 @@ static int dispatch_rw_block_io(struct xen_blkif
*blkif,
>>  				     seg[i].nsec << 9,
>>  				     seg[i].offset) == 0)) {
>>  
>> -			bio = bio_alloc(GFP_KERNEL, nseg-i);
>> +			int nr_iovecs = (nseg-i) > BIO_MAX_PAGES ? BIO_MAX_PAGES :
(nseg-i);
> 
> I''m sure this results in a compiler warning (declaration after
> statement).
No, because it''s the first statement inside the while loop.
> And it surely would read much better if you used some form of
> min() - the shorter line would at once allow you to properly
> blank separate operands from operators.
Sure, I will switch it to min, thanks for the comments.
> 
> Jan
> 
>> +			bio = bio_alloc(GFP_KERNEL, nr_iovecs);
>>  			if (unlikely(bio == NULL))
>>  				goto fail_put_bio;
>>  
> 
>

With the introduction of indirect segments we can receive requests
with a number of segments bigger than the maximum number of allowed
iovecs in a bios, so make sure that blkback doesn't try to allocate a
bios with more iovecs than BIO_MAX_PAGES

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
---
 drivers/block/xen-blkback/blkback.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/drivers/block/xen-blkback/blkback.c
b/drivers/block/xen-blkback/blkback.c
index d622d86..b3897f5d 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -1236,7 +1236,8 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
 				     seg[i].nsec << 9,
 				     seg[i].offset) == 0)) {
 
-			bio = bio_alloc(GFP_KERNEL, nseg-i);
+			int nr_iovecs = min_t(int, (nseg-i), BIO_MAX_PAGES);
+			bio = bio_alloc(GFP_KERNEL, nr_iovecs);
 			if (unlikely(bio == NULL))
 				goto fail_put_bio;
 
-- 
1.7.7.5 (Apple Git-26)


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On Sat, Jun 22, 2013 at 09:59:17AM +0200, Roger Pau Monne
wrote:
> With the introduction of indirect segments we can receive requests
> with a number of segments bigger than the maximum number of allowed
> iovecs in a bios, so make sure that blkback doesn''t try to
allocate a
> bios with more iovecs than BIO_MAX_PAGES
Shouldn''t we just gate the feature-indirect-descriptor value to
take this into account?

What happens if the nseg is > BIO_MAX_PAGES? Do we "lose" the
request
for the remaining segments?
> 
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> ---
>  drivers/block/xen-blkback/blkback.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/block/xen-blkback/blkback.c
b/drivers/block/xen-blkback/blkback.c
> index d622d86..b3897f5d 100644
> --- a/drivers/block/xen-blkback/blkback.c
> +++ b/drivers/block/xen-blkback/blkback.c
> @@ -1236,7 +1236,8 @@ static int dispatch_rw_block_io(struct xen_blkif
*blkif,
>  				     seg[i].nsec << 9,
>  				     seg[i].offset) == 0)) {
>  
> -			bio = bio_alloc(GFP_KERNEL, nseg-i);
> +			int nr_iovecs = min_t(int, (nseg-i), BIO_MAX_PAGES);
> +			bio = bio_alloc(GFP_KERNEL, nr_iovecs);
>  			if (unlikely(bio == NULL))
>  				goto fail_put_bio;
>  
> -- 
> 1.7.7.5 (Apple Git-26)
>

On 24/06/13 15:28, Konrad Rzeszutek Wilk wrote:
> On Sat, Jun 22, 2013 at 09:59:17AM +0200, Roger Pau Monne wrote:
>> With the introduction of indirect segments we can receive requests
>> with a number of segments bigger than the maximum number of allowed
>> iovecs in a bios, so make sure that blkback doesn''t try to
allocate a
>> bios with more iovecs than BIO_MAX_PAGES
> 
> Shouldn''t we just gate the feature-indirect-descriptor value to
> take this into account?
> 
> What happens if the nseg is > BIO_MAX_PAGES? Do we "lose" the
request
> for the remaining segments?
No, we just allocate several bios in order to send the request to the
underlying device.