On 09/01/22 17:08, Eric Blake wrote:
On Wed, Aug 31, 2022 at 09:39:23AM -0500, Eric Blake wrote:
> Since NBD_OPT_LIST_META_CONTEXTS is stateless from the server's point
> of view, there is no reason to make it clear state on the client's
> side. Since we do not otherwise modify h->meta_contexts while
> processing nbd_opt_list_meta_contexts(), we also should not wipe it;
> similarly, while clearing h->exportsize makes sense if we know we are
> going to attempt NBD_OPT_INFO or NBD_OPT_GO, it should not be
> attempted merely because we are doing NBD_OPT_LIST_META_CONTEXTS.
>
> Testsuite coverage for this is in the next patch, for ease of review
> (that is, this is one case where it is easy to swap the patch order to
> see a failure fixed by this patch).
> ---
> generator/states-newstyle-opt-meta-context.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
Given the review comments on 3/12, I'm thinking of squashing in:
diff --git c/generator/states-newstyle-opt-go.c w/generator/states-newstyle-opt-go.c
index 1ca5f099..9b05bec1 100644
--- c/generator/states-newstyle-opt-go.c
+++ w/generator/states-newstyle-opt-go.c
@@ -22,6 +22,7 @@ STATE_MACHINE {
NEWSTYLE.OPT_GO.START:
uint16_t nrinfos = 0;
+ nbd_internal_reset_size_and_flags (h);
if (h->request_block_size)
nrinfos++;
if (h->full_info)
diff --git c/generator/states-newstyle-opt-meta-context.c
w/generator/states-newstyle-opt-meta-context.c
index 3e66b106..6149d61d 100644
--- c/generator/states-newstyle-opt-meta-context.c
+++ w/generator/states-newstyle-opt-meta-context.c
@@ -35,12 +35,12 @@ STATE_MACHINE {
* nbd_opt_list_meta_context()
* -> unconditionally use LIST, next state NEGOTIATING
*
- * For now, we start by unconditionally clearing h->exportsize and friends.
* If SET is conditional, we skip it if h->request_meta is false, if
* structured replies were not negotiated, or if no contexts to request.
* If OPT_GO is later successful, it populates h->exportsize and friends,
* and also sets h->meta_valid if h->request_meta but we skipped SET here.
- * SET then manipulates h->meta_contexts, and sets h->meta_valid on success.
+ * SET then manipulates h->meta_contexts, and sets h->meta_valid on
+ * success, while LIST is stateless.
* There is a callback if and only if the command is unconditional.
*/
assert (h->gflags & LIBNBD_HANDSHAKE_FLAG_FIXED_NEWSTYLE);
@@ -52,7 +52,6 @@ STATE_MACHINE {
else {
assert (CALLBACK_IS_NULL (h->opt_cb.fn.context));
opt = NBD_OPT_SET_META_CONTEXT;
- nbd_internal_reset_size_and_flags (h);
if (h->request_meta) {
for (i = 0; i < h->meta_contexts.len; ++i)
free (h->meta_contexts.ptr[i].name);
I think you may have queued so many changes for v3 around this part of
the code that I feel like skipping this patch for now (in v2).
Laszlo