Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity

Reorder HAVE_BIO_RW_* checks 

The HAVE_BIO_RW_* #ifdef's must appear before REQ_* #ifdef's
in the bio_is_flush() and bio_is_discard() macros.  Linux 2.6.32
era kernels defined both of values and the HAVE_BIO_RW_* must be
used in this case.  This resulted in a panic in zconfig test 5.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Closes #4951
Closes #4959
This commit is contained in:
Brian Behlendorf 2016-08-11 14:58:13 -07:00
parent bea68ec5bf
commit a57228e51c
1 changed files with 12 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
include/linux/blkdev_compat.h
View File

@ -333,6 +333,10 @@ bio_set_flags_failfast(struct block_device *bdev, int *flags)
* and the new preflush behavior introduced in Linux 4.8. This is correct
* in all cases but may have a performance impact for some kernels. It
* has the advantage of minimizing kernel specific changes in the zvol code.
*
* Note that 2.6.32 era kernels provide both BIO_RW_BARRIER and REQ_FLUSH,
* where BIO_RW_BARRIER is the correct interface. Therefore, it is important
* that the HAVE_BIO_RW_BARRIER check occur before the REQ_FLUSH check.
*/
static inline boolean_t
bio_is_flush(struct bio *bio)
@ -343,10 +347,10 @@ bio_is_flush(struct bio *bio)
return (bio->bi_opf & REQ_PREFLUSH);
#elif defined(REQ_PREFLUSH) && !defined(HAVE_BIO_BI_OPF)
return (bio->bi_rw & REQ_PREFLUSH);
#elif defined(REQ_FLUSH)
return (bio->bi_rw & REQ_FLUSH);
#elif defined(HAVE_BIO_RW_BARRIER)
return (bio->bi_rw & (1 << BIO_RW_BARRIER));
#elif defined(REQ_FLUSH)
return (bio->bi_rw & REQ_FLUSH);
#else
#error "Allowing the build will cause flush requests to be ignored. Please "
"file an issue report at: https://github.com/zfsonlinux/zfs/issues/new"
@ -385,16 +389,20 @@ bio_is_fua(struct bio *bio)
*
* In all cases the normal I/O path is used for discards. The only
* difference is how the kernel tags individual I/Os as discards.
*
* Note that 2.6.32 era kernels provide both BIO_RW_DISCARD and REQ_DISCARD,
* where BIO_RW_DISCARD is the correct interface. Therefore, it is important
* that the HAVE_BIO_RW_DISCARD check occur before the REQ_DISCARD check.
*/
static inline boolean_t
bio_is_discard(struct bio *bio)
{
#if defined(HAVE_REQ_OP_DISCARD)
return (bio_op(bio) == REQ_OP_DISCARD);
#elif defined(REQ_DISCARD)
return (bio->bi_rw & REQ_DISCARD);
#elif defined(HAVE_BIO_RW_DISCARD)
return (bio->bi_rw & (1 << BIO_RW_DISCARD));
#elif defined(REQ_DISCARD)
return (bio->bi_rw & REQ_DISCARD);
#else
#error "Allowing the build will cause discard requests to become writes "
"potentially triggering the DMU_MAX_ACCESS assertion. Please file "