| Message ID | 20211021001059.438843-4-jane.chu@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | dax poison recovery with RWF_RECOVERY_DATA flag | expand |
On Wed, Oct 20, 2021 at 06:10:56PM -0600, Jane Chu wrote: > - if (unlikely(is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, > - PFN_PHYS(nr_pages)))) > + if (unlikely(!(flags & DAXDEV_F_RECOVERY) && > + is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, > + PFN_PHYS(nr_pages)))) The indentation here is pretty messed up. Something like this would be move normal: if (unlikely(!(flags & DAXDEV_F_RECOVERY) && is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, PFN_PHYS(nr_pages)))) { but if we don't really need the unlikely we could do an actually readable variant: if (!(flags & DAXDEV_F_RECOVERY) && is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, PFN_PHYS(nr_pages)))
On 10/21/2021 4:23 AM, Christoph Hellwig wrote: > On Wed, Oct 20, 2021 at 06:10:56PM -0600, Jane Chu wrote: >> - if (unlikely(is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, >> - PFN_PHYS(nr_pages)))) >> + if (unlikely(!(flags & DAXDEV_F_RECOVERY) && >> + is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, >> + PFN_PHYS(nr_pages)))) > > The indentation here is pretty messed up. Something like this would > be move normal: > > if (unlikely(!(flags & DAXDEV_F_RECOVERY) && > is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, > PFN_PHYS(nr_pages)))) { > Will do. > but if we don't really need the unlikely we could do an actually > readable variant: > > if (!(flags & DAXDEV_F_RECOVERY) && > is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, PFN_PHYS(nr_pages))) > 'unlikely' is needed because 'RWF_RECOVERY_DATA' flag is not recommended for normal preadv2/pwritev2 usage, it's recommended only if user is aware of or suspect poison in the range. thanks, -jane
diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c index b0b7fd40560e..ed699416655b 100644 --- a/drivers/nvdimm/pmem.c +++ b/drivers/nvdimm/pmem.c @@ -260,8 +260,9 @@ __weak long __pmem_direct_access(struct pmem_device *pmem, pgoff_t pgoff, { resource_size_t offset = PFN_PHYS(pgoff) + pmem->data_offset; - if (unlikely(is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, - PFN_PHYS(nr_pages)))) + if (unlikely(!(flags & DAXDEV_F_RECOVERY) && + is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, + PFN_PHYS(nr_pages)))) return -EIO; if (kaddr)
Let pmem_dax_direct_access() skip badblock checking if the caller intends to do data recovery by providing the DAXDEV_F_RECOVERY flag. Signed-off-by: Jane Chu <jane.chu@oracle.com> --- drivers/nvdimm/pmem.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)