Whamcloud - gitweb
EX-7601 llite: round LDLM lock requests to chunk
authorPatrick Farrell <pfarrell@whamcloud.com>
Wed, 26 Jul 2023 16:34:49 +0000 (12:34 -0400)
committerAndreas Dilger <adilger@whamcloud.com>
Thu, 7 Sep 2023 00:40:35 +0000 (00:40 +0000)
commit8034d86ab4968b85b4a5ceb83a837043332e744d
tree22dde8c92a2454a3fa6fb93d03120394494ac314
parentb429293b9e81a24053429d46a07e6649f65ad040
EX-7601 llite: round LDLM lock requests to chunk

When we do IO with compression, we may need to 'fill' the
compression chunk, reading up pages which have already been
written to storage, so we can compress the whole chunk.

Doing this safely requires that any dlmlock we're using
always covers the full chunk.

The easiest way to do this is to round the entire locking
process to include leading or trailing compression chunks.

Signed-off-by: Patrick Farrell <pfarrell@whamcloud.com>
Change-Id: I3c365844561d0da909e6290f4b58ef2211c2d255
Reviewed-on: https://review.whamcloud.com/c/ex/lustre-release/+/51266
Tested-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
lustre/llite/vvp_io.c