* [PATCH v4 1/5] migration: Refactor error handling in source return path
2023-10-17 20:26 [PATCH v4 0/5] migration: Better error handling in rp thread, allow failures in recover Peter Xu
@ 2023-10-17 20:26 ` Peter Xu
2023-10-19 21:09 ` Fabiano Rosas
2023-10-31 13:47 ` Juan Quintela
2023-10-17 20:26 ` [PATCH v4 2/5] migration: Allow network to fail even during recovery Peter Xu
` (3 subsequent siblings)
4 siblings, 2 replies; 17+ messages in thread
From: Peter Xu @ 2023-10-17 20:26 UTC (permalink / raw)
To: qemu-devel
Cc: Fabiano Rosas, peterx, Philippe Mathieu-Daudé, Juan Quintela
rp_state.error was a boolean used to show error happened in return path
thread. That's not only duplicating error reporting (migrate_set_error),
but also not good enough in that we only do error_report() and set it to
true, we never can keep a history of the exact error and show it in
query-migrate.
To make this better, a few things done:
- Use error_setg() rather than error_report() across the whole lifecycle
of return path thread, keeping the error in an Error*.
- With above, no need to have mark_source_rp_bad(), remove it, alongside
with rp_state.error itself.
- Use migrate_set_error() to apply that captured error to the global
migration object when error occured in this thread.
- Do the same when detected qemufile error in source return path
We need to re-export qemu_file_get_error_obj() to do the last one.
Signed-off-by: Peter Xu <peterx@redhat.com>
---
migration/migration.h | 1 -
migration/qemu-file.h | 1 +
migration/ram.h | 5 +-
migration/migration.c | 121 ++++++++++++++++++-----------------------
migration/qemu-file.c | 2 +-
migration/ram.c | 41 +++++++-------
migration/trace-events | 4 +-
7 files changed, 80 insertions(+), 95 deletions(-)
diff --git a/migration/migration.h b/migration/migration.h
index ae82004892..1d1ac13adb 100644
--- a/migration/migration.h
+++ b/migration/migration.h
@@ -308,7 +308,6 @@ struct MigrationState {
/* Protected by qemu_file_lock */
QEMUFile *from_dst_file;
QemuThread rp_thread;
- bool error;
/*
* We can also check non-zero of rp_thread, but there's no "official"
* way to do this, so this bool makes it slightly more elegant.
diff --git a/migration/qemu-file.h b/migration/qemu-file.h
index a29c37b0d0..25a1a8052f 100644
--- a/migration/qemu-file.h
+++ b/migration/qemu-file.h
@@ -87,6 +87,7 @@ int coroutine_mixed_fn qemu_peek_byte(QEMUFile *f, int offset);
void qemu_file_skip(QEMUFile *f, int size);
int qemu_file_get_error_obj_any(QEMUFile *f1, QEMUFile *f2, Error **errp);
void qemu_file_set_error_obj(QEMUFile *f, int ret, Error *err);
+int qemu_file_get_error_obj(QEMUFile *f, Error **errp);
void qemu_file_set_error(QEMUFile *f, int ret);
int qemu_file_shutdown(QEMUFile *f);
QEMUFile *qemu_file_get_return_path(QEMUFile *f);
diff --git a/migration/ram.h b/migration/ram.h
index 145c915ca7..14ed666d58 100644
--- a/migration/ram.h
+++ b/migration/ram.h
@@ -51,7 +51,8 @@ uint64_t ram_bytes_total(void);
void mig_throttle_counter_reset(void);
uint64_t ram_pagesize_summary(void);
-int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len);
+int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
+ Error **errp);
void ram_postcopy_migrated_memory_release(MigrationState *ms);
/* For outgoing discard bitmap */
void ram_postcopy_send_discard_bitmap(MigrationState *ms);
@@ -71,7 +72,7 @@ void ramblock_recv_bitmap_set(RAMBlock *rb, void *host_addr);
void ramblock_recv_bitmap_set_range(RAMBlock *rb, void *host_addr, size_t nr);
int64_t ramblock_recv_bitmap_send(QEMUFile *file,
const char *block_name);
-int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *rb);
+int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *rb, Error **errp);
bool ramblock_page_is_discarded(RAMBlock *rb, ram_addr_t start);
void postcopy_preempt_shutdown_file(MigrationState *s);
void *postcopy_preempt_thread(void *opaque);
diff --git a/migration/migration.c b/migration/migration.c
index 6ba5e145ac..3123c4a873 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -99,7 +99,7 @@ static int migration_maybe_pause(MigrationState *s,
int *current_active_state,
int new_state);
static void migrate_fd_cancel(MigrationState *s);
-static int close_return_path_on_source(MigrationState *s);
+static bool close_return_path_on_source(MigrationState *s);
static bool migration_needs_multiple_sockets(void)
{
@@ -1430,7 +1430,6 @@ int migrate_init(MigrationState *s, Error **errp)
s->to_dst_file = NULL;
s->state = MIGRATION_STATUS_NONE;
s->rp_state.from_dst_file = NULL;
- s->rp_state.error = false;
s->mbps = 0.0;
s->pages_per_second = 0.0;
s->downtime = 0;
@@ -1755,16 +1754,6 @@ void qmp_migrate_continue(MigrationStatus state, Error **errp)
qemu_sem_post(&s->pause_sem);
}
-/* migration thread support */
-/*
- * Something bad happened to the RP stream, mark an error
- * The caller shall print or trace something to indicate why
- */
-static void mark_source_rp_bad(MigrationState *s)
-{
- s->rp_state.error = true;
-}
-
void migration_rp_wait(MigrationState *s)
{
qemu_sem_wait(&s->rp_state.rp_sem);
@@ -1795,8 +1784,9 @@ static struct rp_cmd_args {
* We're allowed to send more than requested (e.g. to round to our page size)
* and we don't need to send pages that have already been sent.
*/
-static void migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
- ram_addr_t start, size_t len)
+static void
+migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
+ ram_addr_t start, size_t len, Error **errp)
{
long our_host_ps = qemu_real_host_page_size();
@@ -1808,37 +1798,36 @@ static void migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
*/
if (!QEMU_IS_ALIGNED(start, our_host_ps) ||
!QEMU_IS_ALIGNED(len, our_host_ps)) {
- error_report("%s: Misaligned page request, start: " RAM_ADDR_FMT
- " len: %zd", __func__, start, len);
- mark_source_rp_bad(ms);
+ error_setg(errp, "MIG_RP_MSG_REQ_PAGES: Misaligned page request, start:"
+ RAM_ADDR_FMT " len: %zd", start, len);
return;
}
- if (ram_save_queue_pages(rbname, start, len)) {
- mark_source_rp_bad(ms);
- }
+ ram_save_queue_pages(rbname, start, len, errp);
}
-static int migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name)
+static int migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name,
+ Error **errp)
{
RAMBlock *block = qemu_ram_block_by_name(block_name);
if (!block) {
- error_report("%s: invalid block name '%s'", __func__, block_name);
+ error_setg(errp, "MIG_RP_MSG_RECV_BITMAP has invalid block name '%s'",
+ block_name);
return -EINVAL;
}
/* Fetch the received bitmap and refresh the dirty bitmap */
- return ram_dirty_bitmap_reload(s, block);
+ return ram_dirty_bitmap_reload(s, block, errp);
}
-static int migrate_handle_rp_resume_ack(MigrationState *s, uint32_t value)
+static int migrate_handle_rp_resume_ack(MigrationState *s,
+ uint32_t value, Error **errp)
{
trace_source_return_path_thread_resume_ack(value);
if (value != MIGRATION_RESUME_ACK_VALUE) {
- error_report("%s: illegal resume_ack value %"PRIu32,
- __func__, value);
+ error_setg(errp, "illegal resume_ack value %"PRIu32, value);
return -1;
}
@@ -1897,48 +1886,46 @@ static void *source_return_path_thread(void *opaque)
uint32_t tmp32, sibling_error;
ram_addr_t start = 0; /* =0 to silence warning */
size_t len = 0, expected_len;
+ Error *err = NULL;
int res;
trace_source_return_path_thread_entry();
rcu_register_thread();
- while (!ms->rp_state.error && !qemu_file_get_error(rp) &&
- migration_is_setup_or_active(ms->state)) {
+ while (migration_is_setup_or_active(ms->state)) {
trace_source_return_path_thread_loop_top();
+
header_type = qemu_get_be16(rp);
header_len = qemu_get_be16(rp);
if (qemu_file_get_error(rp)) {
- mark_source_rp_bad(ms);
+ qemu_file_get_error_obj(rp, &err);
goto out;
}
if (header_type >= MIG_RP_MSG_MAX ||
header_type == MIG_RP_MSG_INVALID) {
- error_report("RP: Received invalid message 0x%04x length 0x%04x",
- header_type, header_len);
- mark_source_rp_bad(ms);
+ error_setg(&err, "Received invalid message 0x%04x length 0x%04x",
+ header_type, header_len);
goto out;
}
if ((rp_cmd_args[header_type].len != -1 &&
header_len != rp_cmd_args[header_type].len) ||
header_len > sizeof(buf)) {
- error_report("RP: Received '%s' message (0x%04x) with"
- "incorrect length %d expecting %zu",
- rp_cmd_args[header_type].name, header_type, header_len,
- (size_t)rp_cmd_args[header_type].len);
- mark_source_rp_bad(ms);
+ error_setg(&err, "Received '%s' message (0x%04x) with"
+ "incorrect length %d expecting %zu",
+ rp_cmd_args[header_type].name, header_type, header_len,
+ (size_t)rp_cmd_args[header_type].len);
goto out;
}
/* We know we've got a valid header by this point */
res = qemu_get_buffer(rp, buf, header_len);
if (res != header_len) {
- error_report("RP: Failed reading data for message 0x%04x"
- " read %d expected %d",
- header_type, res, header_len);
- mark_source_rp_bad(ms);
+ error_setg(&err, "Failed reading data for message 0x%04x"
+ " read %d expected %d",
+ header_type, res, header_len);
goto out;
}
@@ -1948,8 +1935,7 @@ static void *source_return_path_thread(void *opaque)
sibling_error = ldl_be_p(buf);
trace_source_return_path_thread_shut(sibling_error);
if (sibling_error) {
- error_report("RP: Sibling indicated error %d", sibling_error);
- mark_source_rp_bad(ms);
+ error_setg(&err, "Sibling indicated error %d", sibling_error);
}
/*
* We'll let the main thread deal with closing the RP
@@ -1967,7 +1953,10 @@ static void *source_return_path_thread(void *opaque)
case MIG_RP_MSG_REQ_PAGES:
start = ldq_be_p(buf);
len = ldl_be_p(buf + 8);
- migrate_handle_rp_req_pages(ms, NULL, start, len);
+ migrate_handle_rp_req_pages(ms, NULL, start, len, &err);
+ if (err) {
+ goto out;
+ }
break;
case MIG_RP_MSG_REQ_PAGES_ID:
@@ -1982,32 +1971,32 @@ static void *source_return_path_thread(void *opaque)
expected_len += tmp32;
}
if (header_len != expected_len) {
- error_report("RP: Req_Page_id with length %d expecting %zd",
- header_len, expected_len);
- mark_source_rp_bad(ms);
+ error_setg(&err, "Req_Page_id with length %d expecting %zd",
+ header_len, expected_len);
+ goto out;
+ }
+ migrate_handle_rp_req_pages(ms, (char *)&buf[13], start, len,
+ &err);
+ if (err) {
goto out;
}
- migrate_handle_rp_req_pages(ms, (char *)&buf[13], start, len);
break;
case MIG_RP_MSG_RECV_BITMAP:
if (header_len < 1) {
- error_report("%s: missing block name", __func__);
- mark_source_rp_bad(ms);
+ error_setg(&err, "MIG_RP_MSG_RECV_BITMAP missing block name");
goto out;
}
/* Format: len (1B) + idstr (<255B). This ends the idstr. */
buf[buf[0] + 1] = '\0';
- if (migrate_handle_rp_recv_bitmap(ms, (char *)(buf + 1))) {
- mark_source_rp_bad(ms);
+ if (migrate_handle_rp_recv_bitmap(ms, (char *)(buf + 1), &err)) {
goto out;
}
break;
case MIG_RP_MSG_RESUME_ACK:
tmp32 = ldl_be_p(buf);
- if (migrate_handle_rp_resume_ack(ms, tmp32)) {
- mark_source_rp_bad(ms);
+ if (migrate_handle_rp_resume_ack(ms, tmp32, &err)) {
goto out;
}
break;
@@ -2023,13 +2012,15 @@ static void *source_return_path_thread(void *opaque)
}
out:
- if (qemu_file_get_error(rp)) {
+ if (err) {
+ migrate_set_error(ms, err);
+ error_free(err);
trace_source_return_path_thread_bad_end();
- mark_source_rp_bad(ms);
}
trace_source_return_path_thread_end();
rcu_unregister_thread();
+
return NULL;
}
@@ -2051,12 +2042,11 @@ static int open_return_path_on_source(MigrationState *ms)
return 0;
}
-static int close_return_path_on_source(MigrationState *ms)
+/* Return true if error detected, or false otherwise */
+static bool close_return_path_on_source(MigrationState *ms)
{
- int ret;
-
if (!ms->rp_state.rp_thread_created) {
- return 0;
+ return false;
}
trace_migration_return_path_end_before();
@@ -2074,18 +2064,13 @@ static int close_return_path_on_source(MigrationState *ms)
}
}
- trace_await_return_path_close_on_source_joining();
qemu_thread_join(&ms->rp_state.rp_thread);
ms->rp_state.rp_thread_created = false;
- trace_await_return_path_close_on_source_close();
-
- ret = ms->rp_state.error;
- ms->rp_state.error = false;
-
migration_release_dst_files(ms);
+ trace_migration_return_path_end_after();
- trace_migration_return_path_end_after(ret);
- return ret;
+ /* Return path will persist the error in MigrationState when quit */
+ return migrate_has_error(ms);
}
static inline void
diff --git a/migration/qemu-file.c b/migration/qemu-file.c
index 3fb25148d1..5b8fd8bee1 100644
--- a/migration/qemu-file.c
+++ b/migration/qemu-file.c
@@ -140,7 +140,7 @@ QEMUFile *qemu_file_new_input(QIOChannel *ioc)
*
* If errp is specified, a verbose error message will be copied over.
*/
-static int qemu_file_get_error_obj(QEMUFile *f, Error **errp)
+int qemu_file_get_error_obj(QEMUFile *f, Error **errp)
{
if (!f->last_error) {
return 0;
diff --git a/migration/ram.c b/migration/ram.c
index c844151ee9..544c5b63cf 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -1955,7 +1955,8 @@ static void migration_page_queue_free(RAMState *rs)
* @start: starting address from the start of the RAMBlock
* @len: length (in bytes) to send
*/
-int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len)
+int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
+ Error **errp)
{
RAMBlock *ramblock;
RAMState *rs = ram_state;
@@ -1972,7 +1973,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len)
* Shouldn't happen, we can't reuse the last RAMBlock if
* it's the 1st request.
*/
- error_report("ram_save_queue_pages no previous block");
+ error_setg(errp, "MIG_RP_MSG_REQ_PAGES has no previous block");
return -1;
}
} else {
@@ -1980,16 +1981,17 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len)
if (!ramblock) {
/* We shouldn't be asked for a non-existent RAMBlock */
- error_report("ram_save_queue_pages no block '%s'", rbname);
+ error_setg(errp, "MIG_RP_MSG_REQ_PAGES has no block '%s'", rbname);
return -1;
}
rs->last_req_rb = ramblock;
}
trace_ram_save_queue_pages(ramblock->idstr, start, len);
if (!offset_in_ramblock(ramblock, start + len - 1)) {
- error_report("%s request overrun start=" RAM_ADDR_FMT " len="
- RAM_ADDR_FMT " blocklen=" RAM_ADDR_FMT,
- __func__, start, len, ramblock->used_length);
+ error_setg(errp, "MIG_RP_MSG_REQ_PAGES request overrun, "
+ "start=" RAM_ADDR_FMT " len="
+ RAM_ADDR_FMT " blocklen=" RAM_ADDR_FMT,
+ start, len, ramblock->used_length);
return -1;
}
@@ -2021,9 +2023,9 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len)
assert(len % page_size == 0);
while (len) {
if (ram_save_host_page_urgent(pss)) {
- error_report("%s: ram_save_host_page_urgent() failed: "
- "ramblock=%s, start_addr=0x"RAM_ADDR_FMT,
- __func__, ramblock->idstr, start);
+ error_setg(errp, "ram_save_host_page_urgent() failed: "
+ "ramblock=%s, start_addr=0x"RAM_ADDR_FMT,
+ ramblock->idstr, start);
ret = -1;
break;
}
@@ -4186,7 +4188,7 @@ static int ram_dirty_bitmap_sync_all(MigrationState *s, RAMState *rs)
* This is only used when the postcopy migration is paused but wants
* to resume from a middle point.
*/
-int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block)
+int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
{
int ret = -EINVAL;
/* from_dst_file is always valid because we're within rp_thread */
@@ -4200,8 +4202,8 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block)
trace_ram_dirty_bitmap_reload_begin(block->idstr);
if (s->state != MIGRATION_STATUS_POSTCOPY_RECOVER) {
- error_report("%s: incorrect state %s", __func__,
- MigrationStatus_str(s->state));
+ error_setg(errp, "Reload bitmap in incorrect state %s",
+ MigrationStatus_str(s->state));
return -EINVAL;
}
@@ -4218,9 +4220,8 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block)
/* The size of the bitmap should match with our ramblock */
if (size != local_size) {
- error_report("%s: ramblock '%s' bitmap size mismatch "
- "(0x%"PRIx64" != 0x%"PRIx64")", __func__,
- block->idstr, size, local_size);
+ error_setg(errp, "ramblock '%s' bitmap size mismatch (0x%"PRIx64
+ " != 0x%"PRIx64")", block->idstr, size, local_size);
return -EINVAL;
}
@@ -4229,15 +4230,15 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block)
ret = qemu_file_get_error(file);
if (ret || size != local_size) {
- error_report("%s: read bitmap failed for ramblock '%s': %d"
- " (size 0x%"PRIx64", got: 0x%"PRIx64")",
- __func__, block->idstr, ret, local_size, size);
+ error_setg(errp, "read bitmap failed for ramblock '%s': %d"
+ " (size 0x%"PRIx64", got: 0x%"PRIx64")",
+ block->idstr, ret, local_size, size);
return -EIO;
}
if (end_mark != RAMBLOCK_RECV_BITMAP_ENDING) {
- error_report("%s: ramblock '%s' end mark incorrect: 0x%"PRIx64,
- __func__, block->idstr, end_mark);
+ error_setg(errp, "ramblock '%s' end mark incorrect: 0x%"PRIx64,
+ block->idstr, end_mark);
return -EINVAL;
}
diff --git a/migration/trace-events b/migration/trace-events
index fa9486dffe..d50e8ed87e 100644
--- a/migration/trace-events
+++ b/migration/trace-events
@@ -148,8 +148,6 @@ multifd_tls_outgoing_handshake_complete(void *ioc) "ioc=%p"
multifd_set_outgoing_channel(void *ioc, const char *ioctype, const char *hostname) "ioc=%p ioctype=%s hostname=%s"
# migration.c
-await_return_path_close_on_source_close(void) ""
-await_return_path_close_on_source_joining(void) ""
migrate_set_state(const char *new_state) "new state %s"
migrate_fd_cleanup(void) ""
migrate_fd_error(const char *error_desc) "error=%s"
@@ -166,7 +164,7 @@ migration_completion_postcopy_end_after_complete(void) ""
migration_rate_limit_pre(int ms) "%d ms"
migration_rate_limit_post(int urgent) "urgent: %d"
migration_return_path_end_before(void) ""
-migration_return_path_end_after(int rp_error) "%d"
+migration_return_path_end_after(void) ""
migration_thread_after_loop(void) ""
migration_thread_file_err(void) ""
migration_thread_setup_complete(void) ""
--
2.41.0
^ permalink raw reply related [flat|nested] 17+ messages in thread* Re: [PATCH v4 1/5] migration: Refactor error handling in source return path
2023-10-17 20:26 ` [PATCH v4 1/5] migration: Refactor error handling in source return path Peter Xu
@ 2023-10-19 21:09 ` Fabiano Rosas
2023-10-31 13:47 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Fabiano Rosas @ 2023-10-19 21:09 UTC (permalink / raw)
To: Peter Xu, qemu-devel; +Cc: peterx, Philippe Mathieu-Daudé, Juan Quintela
Peter Xu <peterx@redhat.com> writes:
> rp_state.error was a boolean used to show error happened in return path
> thread. That's not only duplicating error reporting (migrate_set_error),
> but also not good enough in that we only do error_report() and set it to
> true, we never can keep a history of the exact error and show it in
> query-migrate.
>
> To make this better, a few things done:
>
> - Use error_setg() rather than error_report() across the whole lifecycle
> of return path thread, keeping the error in an Error*.
>
> - With above, no need to have mark_source_rp_bad(), remove it, alongside
> with rp_state.error itself.
>
> - Use migrate_set_error() to apply that captured error to the global
> migration object when error occured in this thread.
>
> - Do the same when detected qemufile error in source return path
>
> We need to re-export qemu_file_get_error_obj() to do the last one.
>
> Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [PATCH v4 1/5] migration: Refactor error handling in source return path
2023-10-17 20:26 ` [PATCH v4 1/5] migration: Refactor error handling in source return path Peter Xu
2023-10-19 21:09 ` Fabiano Rosas
@ 2023-10-31 13:47 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Juan Quintela @ 2023-10-31 13:47 UTC (permalink / raw)
To: Peter Xu; +Cc: qemu-devel, Fabiano Rosas, Philippe Mathieu-Daudé
Peter Xu <peterx@redhat.com> wrote:
> rp_state.error was a boolean used to show error happened in return path
> thread. That's not only duplicating error reporting (migrate_set_error),
> but also not good enough in that we only do error_report() and set it to
> true, we never can keep a history of the exact error and show it in
> query-migrate.
>
> To make this better, a few things done:
>
> - Use error_setg() rather than error_report() across the whole lifecycle
> of return path thread, keeping the error in an Error*.
>
> - With above, no need to have mark_source_rp_bad(), remove it, alongside
> with rp_state.error itself.
>
> - Use migrate_set_error() to apply that captured error to the global
> migration object when error occured in this thread.
>
> - Do the same when detected qemufile error in source return path
>
> We need to re-export qemu_file_get_error_obj() to do the last one.
>
> Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Patch is really long.
And you stated in the commit how to split it to make it more "digestable".
^ permalink raw reply [flat|nested] 17+ messages in thread
* [PATCH v4 2/5] migration: Allow network to fail even during recovery
2023-10-17 20:26 [PATCH v4 0/5] migration: Better error handling in rp thread, allow failures in recover Peter Xu
2023-10-17 20:26 ` [PATCH v4 1/5] migration: Refactor error handling in source return path Peter Xu
@ 2023-10-17 20:26 ` Peter Xu
2023-10-31 14:26 ` Juan Quintela
2023-10-17 20:26 ` [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER Peter Xu
` (2 subsequent siblings)
4 siblings, 1 reply; 17+ messages in thread
From: Peter Xu @ 2023-10-17 20:26 UTC (permalink / raw)
To: qemu-devel
Cc: Fabiano Rosas, peterx, Philippe Mathieu-Daudé, Juan Quintela,
Xiaohui Li
Normally the postcopy recover phase should only exist for a super short
period, that's the duration when QEMU is trying to recover from an
interrupted postcopy migration, during which handshake will be carried out
for continuing the procedure with state changes from PAUSED -> RECOVER ->
POSTCOPY_ACTIVE again.
Here RECOVER phase should be super small, that happens right after the
admin specified a new but working network link for QEMU to reconnect to
dest QEMU.
However there can still be case where the channel is broken in this small
RECOVER window.
If it happens, with current code there's no way the src QEMU can got kicked
out of RECOVER stage. No way either to retry the recover in another channel
when established.
This patch allows the RECOVER phase to fail itself too - we're mostly
ready, just some small things missing, e.g. properly kick the main
migration thread out when sleeping on rp_sem when we found that we're at
RECOVER stage. When this happens, it fails the RECOVER itself, and
rollback to PAUSED stage. Then the user can retry another round of
recovery.
To make it even stronger, teach QMP command migrate-pause to explicitly
kick src/dst QEMU out when needed, so even if for some reason the migration
thread didn't got kicked out already by a failing rethrn-path thread, the
admin can also kick it out.
This will be an super, super corner case, but still try to cover that.
One can try to test this with two proxy channels for migration:
(a) socat unix-listen:/tmp/src.sock,reuseaddr,fork tcp:localhost:10000
(b) socat tcp-listen:10000,reuseaddr,fork unix:/tmp/dst.sock
So the migration channel will be:
(a) (b)
src -> /tmp/src.sock -> tcp:10000 -> /tmp/dst.sock -> dst
Then to make QEMU hang at RECOVER stage, one can do below:
(1) stop the postcopy using QMP command postcopy-pause
(2) kill the 2nd proxy (b)
(3) try to recover the postcopy using /tmp/src.sock on src
(4) src QEMU will go into RECOVER stage but won't be able to continue
from there, because the channel is actually broken at (b)
Before this patch, step (4) will make src QEMU stuck in RECOVER stage,
without a way to kick the QEMU out or continue the postcopy again. After
this patch, (4) will quickly fail qemu and bounce back to PAUSED stage.
Admin can also kick QEMU from (4) into PAUSED when needed using
migrate-pause when needed.
After bouncing back to PAUSED stage, one can recover again.
Reported-by: Xiaohui Li <xiaohli@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2111332
Signed-off-by: Peter Xu <peterx@redhat.com>
---
migration/migration.h | 8 ++++--
migration/migration.c | 63 +++++++++++++++++++++++++++++++++++++++----
migration/ram.c | 4 ++-
3 files changed, 67 insertions(+), 8 deletions(-)
diff --git a/migration/migration.h b/migration/migration.h
index 1d1ac13adb..0b695bbfb1 100644
--- a/migration/migration.h
+++ b/migration/migration.h
@@ -494,6 +494,7 @@ int migrate_init(MigrationState *s, Error **errp);
bool migration_is_blocked(Error **errp);
/* True if outgoing migration has entered postcopy phase */
bool migration_in_postcopy(void);
+bool migration_postcopy_is_alive(int state);
MigrationState *migrate_get_current(void);
uint64_t ram_get_total_transferred_pages(void);
@@ -534,8 +535,11 @@ void migration_populate_vfio_info(MigrationInfo *info);
void migration_reset_vfio_bytes_transferred(void);
void postcopy_temp_page_reset(PostcopyTmpPage *tmp_page);
-/* Migration thread waiting for return path thread. */
-void migration_rp_wait(MigrationState *s);
+/*
+ * Migration thread waiting for return path thread. Return non-zero if an
+ * error is detected.
+ */
+int migration_rp_wait(MigrationState *s);
/*
* Kick the migration thread waiting for return path messages. NOTE: the
* name can be slightly confusing (when read as "kick the rp thread"), just
diff --git a/migration/migration.c b/migration/migration.c
index 3123c4a873..0661dad953 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -1348,6 +1348,17 @@ bool migration_in_postcopy(void)
}
}
+bool migration_postcopy_is_alive(int state)
+{
+ switch (state) {
+ case MIGRATION_STATUS_POSTCOPY_ACTIVE:
+ case MIGRATION_STATUS_POSTCOPY_RECOVER:
+ return true;
+ default:
+ return false;
+ }
+}
+
bool migration_in_postcopy_after_devices(MigrationState *s)
{
return migration_in_postcopy() && s->postcopy_after_devices;
@@ -1557,8 +1568,15 @@ void qmp_migrate_pause(Error **errp)
MigrationIncomingState *mis = migration_incoming_get_current();
int ret = 0;
- if (ms->state == MIGRATION_STATUS_POSTCOPY_ACTIVE) {
+ if (migration_postcopy_is_alive(ms->state)) {
/* Source side, during postcopy */
+ Error *error = NULL;
+
+ /* Tell the core migration that we're pausing */
+ error_setg(&error, "Postcopy migration is paused by the user");
+ migrate_set_error(ms, error);
+ error_free(error);
+
qemu_mutex_lock(&ms->qemu_file_lock);
if (ms->to_dst_file) {
ret = qemu_file_shutdown(ms->to_dst_file);
@@ -1567,10 +1585,17 @@ void qmp_migrate_pause(Error **errp)
if (ret) {
error_setg(errp, "Failed to pause source migration");
}
+
+ /*
+ * Kick the migration thread out of any waiting windows (on behalf
+ * of the rp thread).
+ */
+ migration_rp_kick(ms);
+
return;
}
- if (mis->state == MIGRATION_STATUS_POSTCOPY_ACTIVE) {
+ if (migration_postcopy_is_alive(mis->state)) {
ret = qemu_file_shutdown(mis->from_src_file);
if (ret) {
error_setg(errp, "Failed to pause destination migration");
@@ -1579,7 +1604,7 @@ void qmp_migrate_pause(Error **errp)
}
error_setg(errp, "migrate-pause is currently only supported "
- "during postcopy-active state");
+ "during postcopy-active or postcopy-recover state");
}
bool migration_is_blocked(Error **errp)
@@ -1754,9 +1779,21 @@ void qmp_migrate_continue(MigrationStatus state, Error **errp)
qemu_sem_post(&s->pause_sem);
}
-void migration_rp_wait(MigrationState *s)
+int migration_rp_wait(MigrationState *s)
{
+ /* If migration has failure already, ignore the wait */
+ if (migrate_has_error(s)) {
+ return -1;
+ }
+
qemu_sem_wait(&s->rp_state.rp_sem);
+
+ /* After wait, double check that there's no failure */
+ if (migrate_has_error(s)) {
+ return -1;
+ }
+
+ return 0;
}
void migration_rp_kick(MigrationState *s)
@@ -2018,6 +2055,20 @@ out:
trace_source_return_path_thread_bad_end();
}
+ if (ms->state == MIGRATION_STATUS_POSTCOPY_RECOVER) {
+ /*
+ * this will be extremely unlikely: that we got yet another network
+ * issue during recovering of the 1st network failure.. during this
+ * period the main migration thread can be waiting on rp_sem for
+ * this thread to sync with the other side.
+ *
+ * When this happens, explicitly kick the migration thread out of
+ * RECOVER stage and back to PAUSED, so the admin can try
+ * everything again.
+ */
+ migration_rp_kick(ms);
+ }
+
trace_source_return_path_thread_end();
rcu_unregister_thread();
@@ -2482,7 +2533,9 @@ static int postcopy_resume_handshake(MigrationState *s)
qemu_savevm_send_postcopy_resume(s->to_dst_file);
while (s->state == MIGRATION_STATUS_POSTCOPY_RECOVER) {
- migration_rp_wait(s);
+ if (migration_rp_wait(s)) {
+ return -1;
+ }
}
if (s->state == MIGRATION_STATUS_POSTCOPY_ACTIVE) {
diff --git a/migration/ram.c b/migration/ram.c
index 544c5b63cf..973e872284 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -4175,7 +4175,9 @@ static int ram_dirty_bitmap_sync_all(MigrationState *s, RAMState *rs)
/* Wait until all the ramblocks' dirty bitmap synced */
while (qatomic_read(&rs->postcopy_bmap_sync_requested)) {
- migration_rp_wait(s);
+ if (migration_rp_wait(s)) {
+ return -1;
+ }
}
trace_ram_dirty_bitmap_sync_complete();
--
2.41.0
^ permalink raw reply related [flat|nested] 17+ messages in thread* Re: [PATCH v4 2/5] migration: Allow network to fail even during recovery
2023-10-17 20:26 ` [PATCH v4 2/5] migration: Allow network to fail even during recovery Peter Xu
@ 2023-10-31 14:26 ` Juan Quintela
2023-10-31 15:32 ` Peter Xu
0 siblings, 1 reply; 17+ messages in thread
From: Juan Quintela @ 2023-10-31 14:26 UTC (permalink / raw)
To: Peter Xu; +Cc: qemu-devel, Fabiano Rosas, Philippe Mathieu-Daudé,
Xiaohui Li
Peter Xu <peterx@redhat.com> wrote:
> Normally the postcopy recover phase should only exist for a super short
> period, that's the duration when QEMU is trying to recover from an
> interrupted postcopy migration, during which handshake will be carried out
> for continuing the procedure with state changes from PAUSED -> RECOVER ->
> POSTCOPY_ACTIVE again.
>
> Here RECOVER phase should be super small, that happens right after the
> admin specified a new but working network link for QEMU to reconnect to
> dest QEMU.
>
> However there can still be case where the channel is broken in this small
> RECOVER window.
>
> If it happens, with current code there's no way the src QEMU can got kicked
> out of RECOVER stage. No way either to retry the recover in another channel
> when established.
>
> This patch allows the RECOVER phase to fail itself too - we're mostly
> ready, just some small things missing, e.g. properly kick the main
> migration thread out when sleeping on rp_sem when we found that we're at
> RECOVER stage. When this happens, it fails the RECOVER itself, and
> rollback to PAUSED stage. Then the user can retry another round of
> recovery.
>
> To make it even stronger, teach QMP command migrate-pause to explicitly
> kick src/dst QEMU out when needed, so even if for some reason the migration
> thread didn't got kicked out already by a failing rethrn-path thread, the
> admin can also kick it out.
>
> This will be an super, super corner case, but still try to cover that.
>
> One can try to test this with two proxy channels for migration:
>
> (a) socat unix-listen:/tmp/src.sock,reuseaddr,fork tcp:localhost:10000
> (b) socat tcp-listen:10000,reuseaddr,fork unix:/tmp/dst.sock
>
> So the migration channel will be:
>
> (a) (b)
> src -> /tmp/src.sock -> tcp:10000 -> /tmp/dst.sock -> dst
>
> Then to make QEMU hang at RECOVER stage, one can do below:
>
> (1) stop the postcopy using QMP command postcopy-pause
> (2) kill the 2nd proxy (b)
> (3) try to recover the postcopy using /tmp/src.sock on src
> (4) src QEMU will go into RECOVER stage but won't be able to continue
> from there, because the channel is actually broken at (b)
>
> Before this patch, step (4) will make src QEMU stuck in RECOVER stage,
> without a way to kick the QEMU out or continue the postcopy again. After
> this patch, (4) will quickly fail qemu and bounce back to PAUSED stage.
>
> Admin can also kick QEMU from (4) into PAUSED when needed using
> migrate-pause when needed.
>
> After bouncing back to PAUSED stage, one can recover again.
>
> Reported-by: Xiaohui Li <xiaohli@redhat.com>
> Reviewed-by: Fabiano Rosas <farosas@suse.de>
> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2111332
> Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
> -void migration_rp_wait(MigrationState *s)
> +int migration_rp_wait(MigrationState *s)
> {
> + /* If migration has failure already, ignore the wait */
> + if (migrate_has_error(s)) {
> + return -1;
> + }
> +
> qemu_sem_wait(&s->rp_state.rp_sem);
> +
> + /* After wait, double check that there's no failure */
> + if (migrate_has_error(s)) {
> + return -1;
> + }
> +
> + return 0;
> }
Shouldn't this be bool?
We have (too many) functions in migration that returns 0/-1 and set an
error, I think we should change them to return bool. Or even just test
if err is set.
Later, Juan.
^ permalink raw reply [flat|nested] 17+ messages in thread* Re: [PATCH v4 2/5] migration: Allow network to fail even during recovery
2023-10-31 14:26 ` Juan Quintela
@ 2023-10-31 15:32 ` Peter Xu
0 siblings, 0 replies; 17+ messages in thread
From: Peter Xu @ 2023-10-31 15:32 UTC (permalink / raw)
To: Juan Quintela
Cc: qemu-devel, Fabiano Rosas, Philippe Mathieu-Daudé,
Xiaohui Li
On Tue, Oct 31, 2023 at 03:26:42PM +0100, Juan Quintela wrote:
> > -void migration_rp_wait(MigrationState *s)
> > +int migration_rp_wait(MigrationState *s)
> > {
> > + /* If migration has failure already, ignore the wait */
> > + if (migrate_has_error(s)) {
> > + return -1;
> > + }
> > +
> > qemu_sem_wait(&s->rp_state.rp_sem);
> > +
> > + /* After wait, double check that there's no failure */
> > + if (migrate_has_error(s)) {
> > + return -1;
> > + }
> > +
> > + return 0;
> > }
>
> Shouldn't this be bool?
>
> We have (too many) functions in migration that returns 0/-1 and set an
> error, I think we should change them to return bool. Or even just test
> if err is set.
Yeah this patch comes earlier than "switching to bools". I can make them
bool after I rebase to the new pulls and see what's leftover. Thanks.
--
Peter Xu
^ permalink raw reply [flat|nested] 17+ messages in thread
* [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER
2023-10-17 20:26 [PATCH v4 0/5] migration: Better error handling in rp thread, allow failures in recover Peter Xu
2023-10-17 20:26 ` [PATCH v4 1/5] migration: Refactor error handling in source return path Peter Xu
2023-10-17 20:26 ` [PATCH v4 2/5] migration: Allow network to fail even during recovery Peter Xu
@ 2023-10-17 20:26 ` Peter Xu
2023-10-31 22:01 ` Fabiano Rosas
2023-10-17 20:26 ` [PATCH v4 4/5] migration: Change ram_dirty_bitmap_reload() retval to bool Peter Xu
2023-10-17 20:26 ` [PATCH v4 5/5] migration: Change ram_save_queue_pages() " Peter Xu
4 siblings, 1 reply; 17+ messages in thread
From: Peter Xu @ 2023-10-17 20:26 UTC (permalink / raw)
To: qemu-devel
Cc: Fabiano Rosas, peterx, Philippe Mathieu-Daudé, Juan Quintela
From: Fabiano Rosas <farosas@suse.de>
To do so, create two paired sockets, but make them not providing real data.
Feed those fake sockets to src/dst QEMUs for recovery to let them go into
RECOVER stage without going out. Test that we can always kick it out and
recover again with the right ports.
This patch is based on Fabiano's version here:
https://lore.kernel.org/r/877cowmdu0.fsf@suse.de
Signed-off-by: Fabiano Rosas <farosas@suse.de>
[peterx: write commit message, remove case 1, fix bugs, and more]
Signed-off-by: Peter Xu <peterx@redhat.com>
---
tests/qtest/migration-test.c | 102 ++++++++++++++++++++++++++++++++---
1 file changed, 96 insertions(+), 6 deletions(-)
diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c
index e1c110537b..e81d6de000 100644
--- a/tests/qtest/migration-test.c
+++ b/tests/qtest/migration-test.c
@@ -726,6 +726,7 @@ typedef struct {
/* Postcopy specific fields */
void *postcopy_data;
bool postcopy_preempt;
+ bool postcopy_recovery_test_fail;
} MigrateCommon;
static int test_migrate_start(QTestState **from, QTestState **to,
@@ -1379,6 +1380,78 @@ static void test_postcopy_preempt_tls_psk(void)
}
#endif
+static void wait_for_postcopy_status(QTestState *one, const char *status)
+{
+ wait_for_migration_status(one, status,
+ (const char * []) { "failed", "active",
+ "completed", NULL });
+}
+
+static void postcopy_recover_fail(QTestState *from, QTestState *to)
+{
+ int ret, pair1[2], pair2[2];
+ char c;
+
+ /* Create two unrelated socketpairs */
+ ret = qemu_socketpair(PF_LOCAL, SOCK_STREAM, 0, pair1);
+ g_assert_cmpint(ret, ==, 0);
+
+ ret = qemu_socketpair(PF_LOCAL, SOCK_STREAM, 0, pair2);
+ g_assert_cmpint(ret, ==, 0);
+
+ /*
+ * Give the guests unpaired ends of the sockets, so they'll all blocked
+ * at reading. This mimics a wrong channel established.
+ */
+ qtest_qmp_fds_assert_success(from, &pair1[0], 1,
+ "{ 'execute': 'getfd',"
+ " 'arguments': { 'fdname': 'fd-mig' }}");
+ qtest_qmp_fds_assert_success(to, &pair2[0], 1,
+ "{ 'execute': 'getfd',"
+ " 'arguments': { 'fdname': 'fd-mig' }}");
+
+ /*
+ * Write the 1st byte as QEMU_VM_COMMAND (0x8) for the dest socket, to
+ * emulate the 1st byte of a real recovery, but stops from there to
+ * keep dest QEMU in RECOVER. This is needed so that we can kick off
+ * the recover process on dest QEMU (by triggering the G_IO_IN event).
+ *
+ * NOTE: this trick is not needed on src QEMUs, because src doesn't
+ * rely on an pre-existing G_IO_IN event, so it will always trigger the
+ * upcoming recovery anyway even if it can read nothing.
+ */
+#define QEMU_VM_COMMAND 0x08
+ c = QEMU_VM_COMMAND;
+ ret = send(pair2[1], &c, 1, 0);
+ g_assert_cmpint(ret, ==, 1);
+
+ migrate_recover(to, "fd:fd-mig");
+ migrate_qmp(from, "fd:fd-mig", "{'resume': true}");
+
+ /*
+ * Make sure both QEMU instances will go into RECOVER stage, then test
+ * kicking them out using migrate-pause.
+ */
+ wait_for_postcopy_status(from, "postcopy-recover");
+ wait_for_postcopy_status(to, "postcopy-recover");
+
+ /*
+ * This would be issued by the admin upon noticing the hang, we should
+ * make sure we're able to kick this out.
+ */
+ migrate_pause(from);
+ wait_for_postcopy_status(from, "postcopy-paused");
+
+ /* Do the same test on dest */
+ migrate_pause(to);
+ wait_for_postcopy_status(to, "postcopy-paused");
+
+ close(pair1[0]);
+ close(pair1[1]);
+ close(pair2[0]);
+ close(pair2[1]);
+}
+
static void test_postcopy_recovery_common(MigrateCommon *args)
{
QTestState *from, *to;
@@ -1414,9 +1487,17 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
* migrate-recover command can only succeed if destination machine
* is in the paused state
*/
- wait_for_migration_status(to, "postcopy-paused",
- (const char * []) { "failed", "active",
- "completed", NULL });
+ wait_for_postcopy_status(to, "postcopy-paused");
+ wait_for_postcopy_status(from, "postcopy-paused");
+
+ if (args->postcopy_recovery_test_fail) {
+ /*
+ * Test when a wrong socket specified for recover, and then the
+ * ability to kick it out, and continue with a correct socket.
+ */
+ postcopy_recover_fail(from, to);
+ /* continue with a good recovery */
+ }
/*
* Create a new socket to emulate a new channel that is different
@@ -1430,9 +1511,6 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
* Try to rebuild the migration channel using the resume flag and
* the newly created channel
*/
- wait_for_migration_status(from, "postcopy-paused",
- (const char * []) { "failed", "active",
- "completed", NULL });
migrate_qmp(from, uri, "{'resume': true}");
/* Restore the postcopy bandwidth to unlimited */
@@ -1457,6 +1535,15 @@ static void test_postcopy_recovery_compress(void)
test_postcopy_recovery_common(&args);
}
+static void test_postcopy_recovery_double_fail(void)
+{
+ MigrateCommon args = {
+ .postcopy_recovery_test_fail = true,
+ };
+
+ test_postcopy_recovery_common(&args);
+}
+
#ifdef CONFIG_GNUTLS
static void test_postcopy_recovery_tls_psk(void)
{
@@ -3030,6 +3117,9 @@ int main(int argc, char **argv)
qtest_add_func("/migration/postcopy/recovery/compress/plain",
test_postcopy_recovery_compress);
}
+ qtest_add_func("/migration/postcopy/recovery/double-failures",
+ test_postcopy_recovery_double_fail);
+
}
qtest_add_func("/migration/bad_dest", test_baddest);
--
2.41.0
^ permalink raw reply related [flat|nested] 17+ messages in thread* Re: [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER
2023-10-17 20:26 ` [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER Peter Xu
@ 2023-10-31 22:01 ` Fabiano Rosas
2023-10-31 22:19 ` Peter Xu
2023-10-31 22:34 ` Juan Quintela
0 siblings, 2 replies; 17+ messages in thread
From: Fabiano Rosas @ 2023-10-31 22:01 UTC (permalink / raw)
To: Peter Xu, qemu-devel; +Cc: peterx, Philippe Mathieu-Daudé, Juan Quintela
Peter Xu <peterx@redhat.com> writes:
> From: Fabiano Rosas <farosas@suse.de>
>
> To do so, create two paired sockets, but make them not providing real data.
> Feed those fake sockets to src/dst QEMUs for recovery to let them go into
> RECOVER stage without going out. Test that we can always kick it out and
> recover again with the right ports.
>
> This patch is based on Fabiano's version here:
>
> https://lore.kernel.org/r/877cowmdu0.fsf@suse.de
>
> Signed-off-by: Fabiano Rosas <farosas@suse.de>
> [peterx: write commit message, remove case 1, fix bugs, and more]
> Signed-off-by: Peter Xu <peterx@redhat.com>
This patch breaks the windows build. We need this:
-->8--
From 96fee99f2a3c8e11951100d94159eba02dd98540 Mon Sep 17 00:00:00 2001
From: Fabiano Rosas <farosas@suse.de>
Date: Tue, 31 Oct 2023 17:41:56 -0300
Subject: [PATCH] fixup! tests/migration-test: Add a test for postcopy hangs
during RECOVER
Signed-off-by: Fabiano Rosas <farosas@suse.de>
---
tests/qtest/migration-test.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c
index 2359d349cf..604ffe7746 100644
--- a/tests/qtest/migration-test.c
+++ b/tests/qtest/migration-test.c
@@ -1417,6 +1417,7 @@ static void wait_for_postcopy_status(QTestState *one, const char *status)
"completed", NULL });
}
+#ifndef _WIN32
static void postcopy_recover_fail(QTestState *from, QTestState *to)
{
int ret, pair1[2], pair2[2];
@@ -1481,6 +1482,7 @@ static void postcopy_recover_fail(QTestState *from, QTestState *to)
close(pair2[0]);
close(pair2[1]);
}
+#endif
static void test_postcopy_recovery_common(MigrateCommon *args)
{
@@ -1520,6 +1522,7 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
wait_for_postcopy_status(to, "postcopy-paused");
wait_for_postcopy_status(from, "postcopy-paused");
+#ifndef _WIN32
if (args->postcopy_recovery_test_fail) {
/*
* Test when a wrong socket specified for recover, and then the
@@ -1528,6 +1531,7 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
postcopy_recover_fail(from, to);
/* continue with a good recovery */
}
+#endif
/*
* Create a new socket to emulate a new channel that is different
@@ -1565,6 +1569,7 @@ static void test_postcopy_recovery_compress(void)
test_postcopy_recovery_common(&args);
}
+#ifndef _WIN32
static void test_postcopy_recovery_double_fail(void)
{
MigrateCommon args = {
@@ -1573,6 +1578,7 @@ static void test_postcopy_recovery_double_fail(void)
test_postcopy_recovery_common(&args);
}
+#endif
#ifdef CONFIG_GNUTLS
static void test_postcopy_recovery_tls_psk(void)
@@ -3185,9 +3191,10 @@ int main(int argc, char **argv)
qtest_add_func("/migration/postcopy/recovery/compress/plain",
test_postcopy_recovery_compress);
}
+#ifndef _WIN32
qtest_add_func("/migration/postcopy/recovery/double-failures",
test_postcopy_recovery_double_fail);
-
+#endif
}
qtest_add_func("/migration/bad_dest", test_baddest);
--
2.35.3
^ permalink raw reply related [flat|nested] 17+ messages in thread* Re: [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER
2023-10-31 22:01 ` Fabiano Rosas
@ 2023-10-31 22:19 ` Peter Xu
2023-10-31 22:34 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Peter Xu @ 2023-10-31 22:19 UTC (permalink / raw)
To: Fabiano Rosas; +Cc: qemu-devel, Philippe Mathieu-Daudé, Juan Quintela
On Tue, Oct 31, 2023 at 07:01:02PM -0300, Fabiano Rosas wrote:
> This patch breaks the windows build. We need this:
Oops..
I hope this can still catch Juan's next pull, or I'll squash the fix if to
repost.
Thanks,
--
Peter Xu
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER
2023-10-31 22:01 ` Fabiano Rosas
2023-10-31 22:19 ` Peter Xu
@ 2023-10-31 22:34 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Juan Quintela @ 2023-10-31 22:34 UTC (permalink / raw)
To: Fabiano Rosas; +Cc: Peter Xu, qemu-devel, Philippe Mathieu-Daudé
Fabiano Rosas <farosas@suse.de> wrote:
> Peter Xu <peterx@redhat.com> writes:
>
>> From: Fabiano Rosas <farosas@suse.de>
>>
>> To do so, create two paired sockets, but make them not providing real data.
>> Feed those fake sockets to src/dst QEMUs for recovery to let them go into
>> RECOVER stage without going out. Test that we can always kick it out and
>> recover again with the right ports.
>>
>> This patch is based on Fabiano's version here:
>>
>> https://lore.kernel.org/r/877cowmdu0.fsf@suse.de
>>
>> Signed-off-by: Fabiano Rosas <farosas@suse.de>
>> [peterx: write commit message, remove case 1, fix bugs, and more]
>> Signed-off-by: Peter Xu <peterx@redhat.com>
>
> This patch breaks the windows build. We need this:
I was doing this O:-)
migration-next
hit the exact problem that you mention.
Thanks, Juan.
> -->8--
> From 96fee99f2a3c8e11951100d94159eba02dd98540 Mon Sep 17 00:00:00 2001
> From: Fabiano Rosas <farosas@suse.de>
> Date: Tue, 31 Oct 2023 17:41:56 -0300
> Subject: [PATCH] fixup! tests/migration-test: Add a test for postcopy hangs
> during RECOVER
>
> Signed-off-by: Fabiano Rosas <farosas@suse.de>
> ---
> tests/qtest/migration-test.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c
> index 2359d349cf..604ffe7746 100644
> --- a/tests/qtest/migration-test.c
> +++ b/tests/qtest/migration-test.c
> @@ -1417,6 +1417,7 @@ static void wait_for_postcopy_status(QTestState *one, const char *status)
> "completed", NULL });
> }
>
> +#ifndef _WIN32
> static void postcopy_recover_fail(QTestState *from, QTestState *to)
> {
> int ret, pair1[2], pair2[2];
> @@ -1481,6 +1482,7 @@ static void postcopy_recover_fail(QTestState *from, QTestState *to)
> close(pair2[0]);
> close(pair2[1]);
> }
> +#endif
>
> static void test_postcopy_recovery_common(MigrateCommon *args)
> {
> @@ -1520,6 +1522,7 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
> wait_for_postcopy_status(to, "postcopy-paused");
> wait_for_postcopy_status(from, "postcopy-paused");
>
> +#ifndef _WIN32
> if (args->postcopy_recovery_test_fail) {
> /*
> * Test when a wrong socket specified for recover, and then the
> @@ -1528,6 +1531,7 @@ static void test_postcopy_recovery_common(MigrateCommon *args)
> postcopy_recover_fail(from, to);
> /* continue with a good recovery */
> }
> +#endif
>
> /*
> * Create a new socket to emulate a new channel that is different
> @@ -1565,6 +1569,7 @@ static void test_postcopy_recovery_compress(void)
> test_postcopy_recovery_common(&args);
> }
>
> +#ifndef _WIN32
> static void test_postcopy_recovery_double_fail(void)
> {
> MigrateCommon args = {
> @@ -1573,6 +1578,7 @@ static void test_postcopy_recovery_double_fail(void)
>
> test_postcopy_recovery_common(&args);
> }
> +#endif
>
> #ifdef CONFIG_GNUTLS
> static void test_postcopy_recovery_tls_psk(void)
> @@ -3185,9 +3191,10 @@ int main(int argc, char **argv)
> qtest_add_func("/migration/postcopy/recovery/compress/plain",
> test_postcopy_recovery_compress);
> }
> +#ifndef _WIN32
> qtest_add_func("/migration/postcopy/recovery/double-failures",
> test_postcopy_recovery_double_fail);
> -
> +#endif
> }
>
> qtest_add_func("/migration/bad_dest", test_baddest);
^ permalink raw reply [flat|nested] 17+ messages in thread
* [PATCH v4 4/5] migration: Change ram_dirty_bitmap_reload() retval to bool
2023-10-17 20:26 [PATCH v4 0/5] migration: Better error handling in rp thread, allow failures in recover Peter Xu
` (2 preceding siblings ...)
2023-10-17 20:26 ` [PATCH v4 3/5] tests/migration-test: Add a test for postcopy hangs during RECOVER Peter Xu
@ 2023-10-17 20:26 ` Peter Xu
2023-10-19 21:12 ` Fabiano Rosas
2023-10-31 14:30 ` Juan Quintela
2023-10-17 20:26 ` [PATCH v4 5/5] migration: Change ram_save_queue_pages() " Peter Xu
4 siblings, 2 replies; 17+ messages in thread
From: Peter Xu @ 2023-10-17 20:26 UTC (permalink / raw)
To: qemu-devel
Cc: Fabiano Rosas, peterx, Philippe Mathieu-Daudé, Juan Quintela
Now we have a Error** passed into the return path thread stack, which is
even clearer than an int retval. Change ram_dirty_bitmap_reload() and the
callers to use a bool instead to replace errnos.
Suggested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
---
migration/ram.h | 2 +-
migration/migration.c | 18 +++++++++---------
migration/ram.c | 24 ++++++++++++------------
3 files changed, 22 insertions(+), 22 deletions(-)
diff --git a/migration/ram.h b/migration/ram.h
index 14ed666d58..af0290f8ab 100644
--- a/migration/ram.h
+++ b/migration/ram.h
@@ -72,7 +72,7 @@ void ramblock_recv_bitmap_set(RAMBlock *rb, void *host_addr);
void ramblock_recv_bitmap_set_range(RAMBlock *rb, void *host_addr, size_t nr);
int64_t ramblock_recv_bitmap_send(QEMUFile *file,
const char *block_name);
-int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *rb, Error **errp);
+bool ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *rb, Error **errp);
bool ramblock_page_is_discarded(RAMBlock *rb, ram_addr_t start);
void postcopy_preempt_shutdown_file(MigrationState *s);
void *postcopy_preempt_thread(void *opaque);
diff --git a/migration/migration.c b/migration/migration.c
index 0661dad953..dfb8b48dcb 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -1843,29 +1843,29 @@ migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
ram_save_queue_pages(rbname, start, len, errp);
}
-static int migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name,
- Error **errp)
+static bool migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name,
+ Error **errp)
{
RAMBlock *block = qemu_ram_block_by_name(block_name);
if (!block) {
error_setg(errp, "MIG_RP_MSG_RECV_BITMAP has invalid block name '%s'",
block_name);
- return -EINVAL;
+ return false;
}
/* Fetch the received bitmap and refresh the dirty bitmap */
return ram_dirty_bitmap_reload(s, block, errp);
}
-static int migrate_handle_rp_resume_ack(MigrationState *s,
- uint32_t value, Error **errp)
+static bool migrate_handle_rp_resume_ack(MigrationState *s,
+ uint32_t value, Error **errp)
{
trace_source_return_path_thread_resume_ack(value);
if (value != MIGRATION_RESUME_ACK_VALUE) {
error_setg(errp, "illegal resume_ack value %"PRIu32, value);
- return -1;
+ return false;
}
/* Now both sides are active. */
@@ -1875,7 +1875,7 @@ static int migrate_handle_rp_resume_ack(MigrationState *s,
/* Notify send thread that time to continue send pages */
migration_rp_kick(s);
- return 0;
+ return true;
}
/*
@@ -2026,14 +2026,14 @@ static void *source_return_path_thread(void *opaque)
}
/* Format: len (1B) + idstr (<255B). This ends the idstr. */
buf[buf[0] + 1] = '\0';
- if (migrate_handle_rp_recv_bitmap(ms, (char *)(buf + 1), &err)) {
+ if (!migrate_handle_rp_recv_bitmap(ms, (char *)(buf + 1), &err)) {
goto out;
}
break;
case MIG_RP_MSG_RESUME_ACK:
tmp32 = ldl_be_p(buf);
- if (migrate_handle_rp_resume_ack(ms, tmp32, &err)) {
+ if (!migrate_handle_rp_resume_ack(ms, tmp32, &err)) {
goto out;
}
break;
diff --git a/migration/ram.c b/migration/ram.c
index 973e872284..ca77444e18 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -4189,10 +4189,11 @@ static int ram_dirty_bitmap_sync_all(MigrationState *s, RAMState *rs)
* Read the received bitmap, revert it as the initial dirty bitmap.
* This is only used when the postcopy migration is paused but wants
* to resume from a middle point.
+ *
+ * Returns true if succeeded, false for errors.
*/
-int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
+bool ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
{
- int ret = -EINVAL;
/* from_dst_file is always valid because we're within rp_thread */
QEMUFile *file = s->rp_state.from_dst_file;
g_autofree unsigned long *le_bitmap = NULL;
@@ -4206,7 +4207,7 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
if (s->state != MIGRATION_STATUS_POSTCOPY_RECOVER) {
error_setg(errp, "Reload bitmap in incorrect state %s",
MigrationStatus_str(s->state));
- return -EINVAL;
+ return false;
}
/*
@@ -4224,24 +4225,23 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
if (size != local_size) {
error_setg(errp, "ramblock '%s' bitmap size mismatch (0x%"PRIx64
" != 0x%"PRIx64")", block->idstr, size, local_size);
- return -EINVAL;
+ return false;
}
size = qemu_get_buffer(file, (uint8_t *)le_bitmap, local_size);
end_mark = qemu_get_be64(file);
- ret = qemu_file_get_error(file);
- if (ret || size != local_size) {
- error_setg(errp, "read bitmap failed for ramblock '%s': %d"
- " (size 0x%"PRIx64", got: 0x%"PRIx64")",
- block->idstr, ret, local_size, size);
- return -EIO;
+ if (qemu_file_get_error(file) || size != local_size) {
+ error_setg(errp, "read bitmap failed for ramblock '%s': "
+ "(size 0x%"PRIx64", got: 0x%"PRIx64")",
+ block->idstr, local_size, size);
+ return false;
}
if (end_mark != RAMBLOCK_RECV_BITMAP_ENDING) {
error_setg(errp, "ramblock '%s' end mark incorrect: 0x%"PRIx64,
block->idstr, end_mark);
- return -EINVAL;
+ return false;
}
/*
@@ -4273,7 +4273,7 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *block, Error **errp)
*/
migration_rp_kick(s);
- return 0;
+ return true;
}
static int ram_resume_prepare(MigrationState *s, void *opaque)
--
2.41.0
^ permalink raw reply related [flat|nested] 17+ messages in thread* [PATCH v4 5/5] migration: Change ram_save_queue_pages() retval to bool
2023-10-17 20:26 [PATCH v4 0/5] migration: Better error handling in rp thread, allow failures in recover Peter Xu
` (3 preceding siblings ...)
2023-10-17 20:26 ` [PATCH v4 4/5] migration: Change ram_dirty_bitmap_reload() retval to bool Peter Xu
@ 2023-10-17 20:26 ` Peter Xu
2023-10-19 21:13 ` Fabiano Rosas
2023-10-31 14:34 ` Juan Quintela
4 siblings, 2 replies; 17+ messages in thread
From: Peter Xu @ 2023-10-17 20:26 UTC (permalink / raw)
To: qemu-devel
Cc: Fabiano Rosas, peterx, Philippe Mathieu-Daudé, Juan Quintela
After we have errp which contains the more detailed error message, make
ram_save_queue_pages() returns bool in its stack.
Signed-off-by: Peter Xu <peterx@redhat.com>
---
migration/ram.h | 4 ++--
migration/migration.c | 16 ++++++++--------
migration/ram.c | 18 +++++++++---------
3 files changed, 19 insertions(+), 19 deletions(-)
diff --git a/migration/ram.h b/migration/ram.h
index af0290f8ab..e22a6b0d1c 100644
--- a/migration/ram.h
+++ b/migration/ram.h
@@ -51,8 +51,8 @@ uint64_t ram_bytes_total(void);
void mig_throttle_counter_reset(void);
uint64_t ram_pagesize_summary(void);
-int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
- Error **errp);
+bool ram_save_queue_pages(const char *rbname, ram_addr_t start,
+ ram_addr_t len, Error **errp);
void ram_postcopy_migrated_memory_release(MigrationState *ms);
/* For outgoing discard bitmap */
void ram_postcopy_send_discard_bitmap(MigrationState *ms);
diff --git a/migration/migration.c b/migration/migration.c
index dfb8b48dcb..50bf8422c7 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -1820,8 +1820,10 @@ static struct rp_cmd_args {
* Process a request for pages received on the return path,
* We're allowed to send more than requested (e.g. to round to our page size)
* and we don't need to send pages that have already been sent.
+ *
+ * Returns true if succeed, false otherwise.
*/
-static void
+static bool
migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
ram_addr_t start, size_t len, Error **errp)
{
@@ -1837,10 +1839,10 @@ migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
!QEMU_IS_ALIGNED(len, our_host_ps)) {
error_setg(errp, "MIG_RP_MSG_REQ_PAGES: Misaligned page request, start:"
RAM_ADDR_FMT " len: %zd", start, len);
- return;
+ return false;
}
- ram_save_queue_pages(rbname, start, len, errp);
+ return ram_save_queue_pages(rbname, start, len, errp);
}
static bool migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name,
@@ -1990,8 +1992,7 @@ static void *source_return_path_thread(void *opaque)
case MIG_RP_MSG_REQ_PAGES:
start = ldq_be_p(buf);
len = ldl_be_p(buf + 8);
- migrate_handle_rp_req_pages(ms, NULL, start, len, &err);
- if (err) {
+ if (!migrate_handle_rp_req_pages(ms, NULL, start, len, &err)) {
goto out;
}
break;
@@ -2012,9 +2013,8 @@ static void *source_return_path_thread(void *opaque)
header_len, expected_len);
goto out;
}
- migrate_handle_rp_req_pages(ms, (char *)&buf[13], start, len,
- &err);
- if (err) {
+ if (!migrate_handle_rp_req_pages(ms, (char *)&buf[13], start, len,
+ &err)) {
goto out;
}
break;
diff --git a/migration/ram.c b/migration/ram.c
index ca77444e18..aca9ae5846 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -1948,15 +1948,15 @@ static void migration_page_queue_free(RAMState *rs)
*
* A request from postcopy destination for example.
*
- * Returns zero on success or negative on error
+ * Returns true on success or false on error (detailed error put in @errp)
*
* @rbname: Name of the RAMBLock of the request. NULL means the
* same that last one.
* @start: starting address from the start of the RAMBlock
* @len: length (in bytes) to send
*/
-int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
- Error **errp)
+bool ram_save_queue_pages(const char *rbname, ram_addr_t start,
+ ram_addr_t len, Error **errp)
{
RAMBlock *ramblock;
RAMState *rs = ram_state;
@@ -1974,7 +1974,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
* it's the 1st request.
*/
error_setg(errp, "MIG_RP_MSG_REQ_PAGES has no previous block");
- return -1;
+ return false;
}
} else {
ramblock = qemu_ram_block_by_name(rbname);
@@ -1982,7 +1982,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
if (!ramblock) {
/* We shouldn't be asked for a non-existent RAMBlock */
error_setg(errp, "MIG_RP_MSG_REQ_PAGES has no block '%s'", rbname);
- return -1;
+ return false;
}
rs->last_req_rb = ramblock;
}
@@ -1992,7 +1992,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
"start=" RAM_ADDR_FMT " len="
RAM_ADDR_FMT " blocklen=" RAM_ADDR_FMT,
start, len, ramblock->used_length);
- return -1;
+ return false;
}
/*
@@ -2003,7 +2003,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
ram_addr_t page_start = start >> TARGET_PAGE_BITS;
size_t page_size = qemu_ram_pagesize(ramblock);
PageSearchStatus *pss = &ram_state->pss[RAM_CHANNEL_POSTCOPY];
- int ret = 0;
+ bool ret = true;
qemu_mutex_lock(&rs->bitmap_mutex);
@@ -2026,7 +2026,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
error_setg(errp, "ram_save_host_page_urgent() failed: "
"ramblock=%s, start_addr=0x"RAM_ADDR_FMT,
ramblock->idstr, start);
- ret = -1;
+ ret = false;
break;
}
/*
@@ -2057,7 +2057,7 @@ int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len,
migration_make_urgent_request();
qemu_mutex_unlock(&rs->src_page_req_mutex);
- return 0;
+ return true;
}
static bool save_page_use_compression(RAMState *rs)
--
2.41.0
^ permalink raw reply related [flat|nested] 17+ messages in thread* Re: [PATCH v4 5/5] migration: Change ram_save_queue_pages() retval to bool
2023-10-17 20:26 ` [PATCH v4 5/5] migration: Change ram_save_queue_pages() " Peter Xu
@ 2023-10-19 21:13 ` Fabiano Rosas
2023-10-31 14:34 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Fabiano Rosas @ 2023-10-19 21:13 UTC (permalink / raw)
To: Peter Xu, qemu-devel; +Cc: peterx, Philippe Mathieu-Daudé, Juan Quintela
Peter Xu <peterx@redhat.com> writes:
> After we have errp which contains the more detailed error message, make
> ram_save_queue_pages() returns bool in its stack.
>
> Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [PATCH v4 5/5] migration: Change ram_save_queue_pages() retval to bool
2023-10-17 20:26 ` [PATCH v4 5/5] migration: Change ram_save_queue_pages() " Peter Xu
2023-10-19 21:13 ` Fabiano Rosas
@ 2023-10-31 14:34 ` Juan Quintela
1 sibling, 0 replies; 17+ messages in thread
From: Juan Quintela @ 2023-10-31 14:34 UTC (permalink / raw)
To: Peter Xu; +Cc: qemu-devel, Fabiano Rosas, Philippe Mathieu-Daudé
Peter Xu <peterx@redhat.com> wrote:
> After we have errp which contains the more detailed error message, make
> ram_save_queue_pages() returns bool in its stack.
>
> Signed-off-by: Peter Xu <peterx@redhat.com>
> ---
> migration/ram.h | 4 ++--
> migration/migration.c | 16 ++++++++--------
> migration/ram.c | 18 +++++++++---------
> 3 files changed, 19 insertions(+), 19 deletions(-)
>
> static bool migrate_handle_rp_recv_bitmap(MigrationState *s, char *block_name,
> @@ -1990,8 +1992,7 @@ static void *source_return_path_thread(void *opaque)
> case MIG_RP_MSG_REQ_PAGES:
> start = ldq_be_p(buf);
> len = ldl_be_p(buf + 8);
> - migrate_handle_rp_req_pages(ms, NULL, start, len, &err);
> - if (err) {
> + if (!migrate_handle_rp_req_pages(ms, NULL, start, len, &err)) {
I thought we were headed the other direction.
Make functions return void and just check if the error is set.
I think just change the type form returning int to return void will make
the trick.
Later, Juan.
^ permalink raw reply [flat|nested] 17+ messages in thread