osd: clear data digest when write_trunc

like CEPH_OSD_OP_TRUNCATE, the truncate within CEPH_OSD_OP_WRITE should
clear oi.data_digest too.

consider such circumstance:
[write 0~10]
[write 5~10 [1@5]]

the first write will generate data_digest, if the second write don't
clear digest, it will continue calc digest and make crc mismatch.

Fixes: https://tracker.ceph.com/issues/50763
Signed-off-by: Zengran Zhang <zhangzengran@sangfor.com.cn>
(cherry picked from commit 0e4ffe5e51291d1776aacfc0d88aa3b5f4c3ecca)

diff --git a/src/osd/PrimaryLogPG.cc b/src/osd/PrimaryLogPG.cc
index 71f5e68ab1b2cd58416c0893052934e2b9aed82c..9fb22e0f9ee531e579a83f3331d5e4089c75cb4e 100644 (file)
--- a/src/osd/PrimaryLogPG.cc
+++ b/src/osd/PrimaryLogPG.cc
@@ -6559,6 +6559,7 @@ int PrimaryLogPG::do_osd_ops(OpContext *ctx, vector<OSDOp>& ops)
                oi.size - op.extent.truncate_size);
              ctx->modified_ranges.union_of(trim);
              ctx->clean_regions.mark_data_region_dirty(op.extent.truncate_size, oi.size - op.extent.truncate_size);
+             oi.clear_data_digest();
            }
            if (op.extent.truncate_size != oi.size) {
               truncate_update_size_and_usage(ctx->delta_stats,