[PATCH 02/11] verify: fix verify_offset when used with pattern_hdr

Vincent Fu <[email protected]>
Newsgroups org.kernel.vger.fio
Message-ID <[email protected]>
When using verify=pattern_hdr with verify_offset, we cannot reuse the
data already in a buffer because some of the pattern bytes have been
swapped with the verify header. Trying to reuse the buffer contents just
results in the header being swapped back and forth between the
verify_offset location and the beginning of the verify_interval.

Fix this by avoiding reuse of buffer contents.

Failing test case example:
root@localhost:~/fio-dev/fio-canonical# ./fio --name=verify --filesize=8192 --verify_offset=1024 --verify_pattern=1 --rw=write
verify: (g=0): rw=write, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=psync, iodepth=1
fio-3.39-38-gf18c
Starting 1 process
fio: got pattern 'ca', wanted '01'. Bad bits 5
fio: bad pattern block offset 1024
pattern: verify failed at file verify.0.0 offset 4096, length 4096 (requested block: offset=4096, length=4096, flags=88)
fio: pid=3624903, err=84/file:io_u.c:2258, func=io_u_sync_complete, error=Invalid or incomplete multibyte or wide character

verify: (groupid=0, jobs=1): err=84 (file:io_u.c:2258, func=io_u_sync_complete, error=Invalid or incomplete multibyte or wide character): pid=3624903: Fri Apr 11 22:56:01 2025
  read: IOPS=2000, BW=8000KiB/s (8192kB/s)(8192B/1msec)
    clat (nsec): min=2185, max=15807, avg=8996.00, stdev=9632.21
     lat (nsec): min=2280, max=16120, avg=9200.00, stdev=9786.36
    clat percentiles (nsec):
     |  1.00th=[ 2192],  5.00th=[ 2192], 10.00th=[ 2192], 20.00th=[ 2192],
     | 30.00th=[ 2192], 40.00th=[ 2192], 50.00th=[ 2192], 60.00th=[15808],
     | 70.00th=[15808], 80.00th=[15808], 90.00th=[15808], 95.00th=[15808],
     | 99.00th=[15808], 99.50th=[15808], 99.90th=[15808], 99.95th=[15808],
     | 99.99th=[15808]
  write: IOPS=2000, BW=8000KiB/s (8192kB/s)(8192B/1msec); 0 zone resets
    clat (nsec): min=7910, max=82940, avg=45425.00, stdev=53054.22
     lat (usec): min=9, max=100, avg=55.03, stdev=64.49
    clat percentiles (nsec):
     |  1.00th=[ 7904],  5.00th=[ 7904], 10.00th=[ 7904], 20.00th=[ 7904],
     | 30.00th=[ 7904], 40.00th=[ 7904], 50.00th=[ 7904], 60.00th=[82432],
     | 70.00th=[82432], 80.00th=[82432], 90.00th=[82432], 95.00th=[82432],
     | 99.00th=[82432], 99.50th=[82432], 99.90th=[82432], 99.95th=[82432],
     | 99.99th=[82432]
  lat (usec)   : 4=25.00%, 10=25.00%, 20=25.00%, 100=25.00%
  cpu          : usr=0.00%, sys=0.00%, ctx=0, majf=0, minf=19
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=20.0%, 4=80.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=2,2,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
   READ: bw=8000KiB/s (8192kB/s), 8000KiB/s-8000KiB/s (8192kB/s-8192kB/s), io=8192B (8192B), run=1-1msec
  WRITE: bw=8000KiB/s (8192kB/s), 8000KiB/s-8000KiB/s (8192kB/s-8192kB/s), io=8192B (8192B), run=1-1msec

Disk stats (read/write):
  sda: ios=0/0, sectors=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%

Signed-off-by: Vincent Fu <[email protected]>
---
 verify.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/verify.c b/verify.c
index 928bdd54..e7e619d3 100644
--- a/verify.c
+++ b/verify.c
@@ -65,9 +65,10 @@ void fill_verify_pattern(struct thread_data *td, void *p, unsigned int len,
 		return;
 	}
 
-	/* Skip if we were here and we do not need to patch pattern
-	 * with format */
-	if (!td->o.verify_fmt_sz && io_u->buf_filled_len >= len) {
+	/* Skip if we were here and we do not need to patch pattern with
+	 * format. However, we cannot skip if verify_offset is set because we
+	 * have swapped the header with pattern bytes */
+	if (!td->o.verify_fmt_sz && io_u->buf_filled_len >= len && !td->o.verify_offset) {
 		dprint(FD_VERIFY, "using already filled verify pattern b=%d len=%u\n",
 			o->verify_pattern_bytes, len);
 		return;
-- 
2.47.2
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.