summaryrefslogtreecommitdiff
path: root/fs/exofs
diff options
context:
space:
mode:
authorBoaz Harrosh <bharrosh@panasas.com>2013-11-21 17:58:08 +0200
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>2014-02-13 13:48:00 -0800
commited6161470c6c8c45d97211d2426eb050dc382a4a (patch)
tree9e4c68894235b47f10835930c8bee8ba80aad269 /fs/exofs
parentc9f36c5a61a49fd873fc38aa447934ef2417d1d8 (diff)
downloadlinux-3.10-ed6161470c6c8c45d97211d2426eb050dc382a4a.tar.gz
linux-3.10-ed6161470c6c8c45d97211d2426eb050dc382a4a.tar.bz2
linux-3.10-ed6161470c6c8c45d97211d2426eb050dc382a4a.zip
ore: Fix wrong math in allocation of per device BIO
commit aad560b7f63b495f48a7232fd086c5913a676e6f upstream. At IO preparation we calculate the max pages at each device and allocate a BIO per device of that size. The calculation was wrong on some unaligned corner cases offset/length combination and would make prepare return with -ENOMEM. This would be bad for pnfs-objects that would in that case IO through MDS. And fatal for exofs were it would fail writes with EIO. Fix it by doing the proper math, that will work in all cases. (I ran a test with all possible offset/length combinations this time round). Also when reading we do not need to allocate for the parity units since we jump over them. Also lower the max_io_length to take into account the parity pages so not to allocate BIOs bigger than PAGE_SIZE Signed-off-by: Boaz Harrosh <bharrosh@panasas.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Diffstat (limited to 'fs/exofs')
-rw-r--r--fs/exofs/ore.c37
1 files changed, 25 insertions, 12 deletions
diff --git a/fs/exofs/ore.c b/fs/exofs/ore.c
index b7442288860..85cde3e7629 100644
--- a/fs/exofs/ore.c
+++ b/fs/exofs/ore.c
@@ -103,7 +103,7 @@ int ore_verify_layout(unsigned total_comps, struct ore_layout *layout)
layout->max_io_length =
(BIO_MAX_PAGES_KMALLOC * PAGE_SIZE - layout->stripe_unit) *
- layout->group_width;
+ (layout->group_width - layout->parity);
if (layout->parity) {
unsigned stripe_length =
(layout->group_width - layout->parity) *
@@ -286,7 +286,8 @@ int ore_get_rw_state(struct ore_layout *layout, struct ore_components *oc,
if (length) {
ore_calc_stripe_info(layout, offset, length, &ios->si);
ios->length = ios->si.length;
- ios->nr_pages = (ios->length + PAGE_SIZE - 1) / PAGE_SIZE;
+ ios->nr_pages = ((ios->offset & (PAGE_SIZE - 1)) +
+ ios->length + PAGE_SIZE - 1) / PAGE_SIZE;
if (layout->parity)
_ore_post_alloc_raid_stuff(ios);
}
@@ -536,6 +537,7 @@ void ore_calc_stripe_info(struct ore_layout *layout, u64 file_offset,
u64 H = LmodS - G * T;
u32 N = div_u64(H, U);
+ u32 Nlast;
/* "H - (N * U)" is just "H % U" so it's bound to u32 */
u32 C = (u32)(H - (N * U)) / stripe_unit + G * group_width;
@@ -568,6 +570,10 @@ void ore_calc_stripe_info(struct ore_layout *layout, u64 file_offset,
si->length = T - H;
if (si->length > length)
si->length = length;
+
+ Nlast = div_u64(H + si->length + U - 1, U);
+ si->maxdevUnits = Nlast - N;
+
si->M = M;
}
EXPORT_SYMBOL(ore_calc_stripe_info);
@@ -583,13 +589,16 @@ int _ore_add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg,
int ret;
if (per_dev->bio == NULL) {
- unsigned pages_in_stripe = ios->layout->group_width *
- (ios->layout->stripe_unit / PAGE_SIZE);
- unsigned nr_pages = ios->nr_pages * ios->layout->group_width /
- (ios->layout->group_width -
- ios->layout->parity);
- unsigned bio_size = (nr_pages + pages_in_stripe) /
- ios->layout->group_width;
+ unsigned bio_size;
+
+ if (!ios->reading) {
+ bio_size = ios->si.maxdevUnits;
+ } else {
+ bio_size = (ios->si.maxdevUnits + 1) *
+ (ios->layout->group_width - ios->layout->parity) /
+ ios->layout->group_width;
+ }
+ bio_size *= (ios->layout->stripe_unit / PAGE_SIZE);
per_dev->bio = bio_kmalloc(GFP_KERNEL, bio_size);
if (unlikely(!per_dev->bio)) {
@@ -609,8 +618,12 @@ int _ore_add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg,
added_len = bio_add_pc_page(q, per_dev->bio, pages[pg],
pglen, pgbase);
if (unlikely(pglen != added_len)) {
- ORE_DBGMSG("Failed bio_add_pc_page bi_vcnt=%u\n",
- per_dev->bio->bi_vcnt);
+ /* If bi_vcnt == bi_max then this is a SW BUG */
+ ORE_DBGMSG("Failed bio_add_pc_page bi_vcnt=0x%x "
+ "bi_max=0x%x BIO_MAX=0x%x cur_len=0x%x\n",
+ per_dev->bio->bi_vcnt,
+ per_dev->bio->bi_max_vecs,
+ BIO_MAX_PAGES_KMALLOC, cur_len);
ret = -ENOMEM;
goto out;
}
@@ -1098,7 +1111,7 @@ int ore_truncate(struct ore_layout *layout, struct ore_components *oc,
size_attr->attr = g_attr_logical_length;
size_attr->attr.val_ptr = &size_attr->newsize;
- ORE_DBGMSG("trunc(0x%llx) obj_offset=0x%llx dev=%d\n",
+ ORE_DBGMSG2("trunc(0x%llx) obj_offset=0x%llx dev=%d\n",
_LLU(oc->comps->obj.id), _LLU(obj_size), i);
ret = _truncate_mirrors(ios, i * ios->layout->mirrors_p1,
&size_attr->attr);