Revert "libnvdimm: band aid btt vs clear poison locking"
authorDan Williams <dan.j.williams@intel.com>
Thu, 13 Apr 2017 21:37:39 +0000 (14:37 -0700)
committerDan Williams <dan.j.williams@intel.com>
Fri, 14 Apr 2017 20:29:01 +0000 (13:29 -0700)
This reverts commit 4aa5615e080a "libnvdimm: band aid btt vs clear
poison locking".

Now that poison list locking has been converted to a spinlock and poison
list entry allocation during i/o has been converted to GFP_NOWAIT,
revert the band-aid that disabled error clearing from btt i/o.

Cc: Vishal Verma <vishal.l.verma@intel.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/nvdimm/claim.c

index ca6d572c48fcb62d7c8a83e4874f18b625caaeb8..b3323c0697f6239ebbfe757137cde8352fe3c480 100644 (file)
@@ -243,15 +243,7 @@ static int nsio_rw_bytes(struct nd_namespace_common *ndns,
        }
 
        if (unlikely(is_bad_pmem(&nsio->bb, sector, sz_align))) {
-               /*
-                * FIXME: nsio_rw_bytes() may be called from atomic
-                * context in the btt case and nvdimm_clear_poison()
-                * takes a sleeping lock. Until the locking can be
-                * reworked this capability requires that the namespace
-                * is not claimed by btt.
-                */
-               if (IS_ALIGNED(offset, 512) && IS_ALIGNED(size, 512)
-                               && (!ndns->claim || !is_nd_btt(ndns->claim))) {
+               if (IS_ALIGNED(offset, 512) && IS_ALIGNED(size, 512)) {
                        long cleared;
 
                        cleared = nvdimm_clear_poison(&ndns->dev, offset, size);