lib/ftl: Fix I/O alignment in dirty shutdown restore

Changed to use 4k alignment in dirty shutdown I/Os. Otherwise the
scatter gather lists used in QEMU for underlying file/block device
would use an extra entry (e.g. 17 for 16 sector writes), and eventually
some I/Os would write to offset 0 in underlying file, corrupting head
metadata.

Signed-off-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Change-Id: If8c88ce708529b094a09c8ee952912cc22cd53b9
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458090
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This commit is contained in:
Mateusz Kozlowski 2019-06-14 16:24:42 +02:00 committed by Darek Stojaczyk
parent 67d027ece9
commit 38bb4bcdee

View File

@ -180,7 +180,6 @@ ftl_restore_check_seq(const struct ftl_restore *restore)
for (i = 0; i < ftl_dev_num_bands(dev); ++i) {
rband = &restore->bands[i];
if (rband->md_status != FTL_MD_SUCCESS) {
continue;
}
@ -533,7 +532,14 @@ ftl_restore_pad_band(struct ftl_restore_band *rband)
ppa = band->chunk_buf[i].start_ppa;
ppa.lbk = info.wp;
buffer = spdk_dma_zmalloc(FTL_BLOCK_SIZE * dev->xfer_size, sizeof(uint32_t), NULL);
/*
* We need 4k alignment for lightnvm writes; otherwise, due to a bug in QEMU,
* scatter gather lists to underlying block device become broken as the number of
* incoming offsets (dev->xfer_size) would be smaller than the calculated size of
* sgl (dev->xfer_size+1), which eventually results in some part of the write
* hitting offset 0 of the drive, overwriting head_md of band 0.
*/
buffer = spdk_dma_zmalloc(FTL_BLOCK_SIZE * dev->xfer_size, FTL_BLOCK_SIZE, NULL);
if (spdk_unlikely(!buffer)) {
rc = -ENOMEM;
goto error;