When buffered writes are used then two things change:
1 - signals will propogate to the submit() call (as they would be
effectively ignored when going async)
2 - CQE ordering will change
Fix the read-write for both of these cases by ignoring the signal and
handling CQE ordering.
Signed-off-by: Dylan Yudaken <dylany@xxxxxx>
---
test/read-write.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/test/read-write.c b/test/read-write.c
index c5cc469f258d..f50e8242afa4 100644
--- a/test/read-write.c
+++ b/test/read-write.c
@@ -759,6 +759,7 @@ static int test_write_efbig(void)
goto err;
}
io_uring_prep_writev(sqe, fd, &vecs[i], 1, off);
+ io_uring_sqe_set_data64(sqe, i);
off += BS;
}
@@ -774,7 +775,7 @@ static int test_write_efbig(void)
fprintf(stderr, "wait_cqe=%d\n", ret);
goto err;
}
- if (i < 16) {
+ if (cqe->user_data < 16) {
if (cqe->res != BS) {
fprintf(stderr, "bad write: %d\n", cqe->res);
goto err;
@@ -819,6 +820,8 @@ int main(int argc, char *argv[])
t_create_file(fname, FILE_SIZE);
}
+ signal(SIGXFSZ, SIG_IGN);
+
vecs = t_create_buffers(BUFFERS, BS);
/* if we don't have nonvec read, skip testing that */
base-commit: 15cc446cb8b0caa5c939423d31182eba66141539
--
2.30.2
