* [PATCH] fstests: log-writes: Add new option to replay/find next write to sector
@ 2019-03-28 5:35 Qu Wenruo
2019-04-06 12:10 ` Eryu Guan
2019-04-06 16:49 ` Amir Goldstein
0 siblings, 2 replies; 3+ messages in thread
From: Qu Wenruo @ 2019-03-28 5:35 UTC (permalink / raw)
To: linux-btrfs, fstests
For kernel operation we have METADATA/FUA/FLUSH flags as a beacon, but
for user space write we don't have any useful flag unless the user space
tool call fsync() to generate a FLUSH bio.
This means for user space write, we don't really have an equivalent of
--next-fua to find super block write.
So this patch will add a new option, --next-write <sector> to
replay/find next write to certain sector.
And normally the <sector> should be super block sector number.
With that, we can replay to super block write even it's user space
triggering the write.
Signed-off-by: Qu Wenruo <wqu@suse.com>
---
src/log-writes/replay-log.c | 16 +++++++++++++++-
1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/src/log-writes/replay-log.c b/src/log-writes/replay-log.c
index c16df40e5741..0d0ec49c8a00 100644
--- a/src/log-writes/replay-log.c
+++ b/src/log-writes/replay-log.c
@@ -9,6 +9,7 @@
enum option_indexes {
NEXT_FLUSH,
NEXT_FUA,
+ NEXT_WRITE,
START_ENTRY,
END_MARK,
LOG,
@@ -28,6 +29,7 @@ enum option_indexes {
static struct option long_options[] = {
{"next-flush", no_argument, NULL, 0},
{"next-fua", no_argument, NULL, 0},
+ {"next-write", required_argument, NULL, 0},
{"start-entry", required_argument, NULL, 0},
{"end-mark", required_argument, NULL, 0},
{"log", required_argument, NULL, 0},
@@ -53,6 +55,8 @@ static void usage(void)
fprintf(stderr, "\t--limit <number> - number of entries to replay\n");
fprintf(stderr, "\t--next-flush - replay to/find the next flush\n");
fprintf(stderr, "\t--next-fua - replay to/find the next fua\n");
+ fprintf(stderr, "\t--next-write <sector> - replay to/find the next "
+ "write to <sector>\n");
fprintf(stderr, "\t--start-entry <entry> - start at the given "
"entry #\n");
fprintf(stderr, "\t--start-mark <mark> - mark to start from\n");
@@ -139,6 +143,7 @@ int main(int argc, char **argv)
char *logfile = NULL, *replayfile = NULL, *fsck_command = NULL;
struct log_write_entry *entry;
u64 stop_flags = 0;
+ u64 stop_sector = 0;
u64 start_entry = 0;
u64 start_sector = 0;
u64 end_sector = -1ULL;
@@ -173,6 +178,14 @@ int main(int argc, char **argv)
case NEXT_FUA:
stop_flags |= LOG_FUA_FLAG;
break;
+ case NEXT_WRITE:
+ stop_sector = strtoull(optarg, &tmp, 0);
+ if (tmp && *tmp != '\0') {
+ fprintf(stderr, "Invalid sector number\n");
+ exit(1);
+ }
+ tmp = NULL;
+ break;
case START_ENTRY:
start_entry = strtoull(optarg, &tmp, 0);
if (tmp && *tmp != '\0') {
@@ -324,7 +337,8 @@ int main(int argc, char **argv)
while ((ret = log_seek_next_entry(log, entry, 1)) == 0) {
num_entries++;
if ((run_limit && num_entries == run_limit) ||
- should_stop(entry, stop_flags, end_mark)) {
+ should_stop(entry, stop_flags, end_mark) ||
+ (stop_sector && entry->sector == stop_sector)) {
printf("%llu@%llu\n",
(unsigned long long)log->cur_entry - 1,
log->cur_pos / log->sectorsize);
--
2.21.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH] fstests: log-writes: Add new option to replay/find next write to sector
2019-03-28 5:35 [PATCH] fstests: log-writes: Add new option to replay/find next write to sector Qu Wenruo
@ 2019-04-06 12:10 ` Eryu Guan
2019-04-06 16:49 ` Amir Goldstein
1 sibling, 0 replies; 3+ messages in thread
From: Eryu Guan @ 2019-04-06 12:10 UTC (permalink / raw)
To: Qu Wenruo; +Cc: linux-btrfs, fstests, Josef Bacik, Amir Goldstein
On Thu, Mar 28, 2019 at 01:35:07PM +0800, Qu Wenruo wrote:
> For kernel operation we have METADATA/FUA/FLUSH flags as a beacon, but
> for user space write we don't have any useful flag unless the user space
> tool call fsync() to generate a FLUSH bio.
>
> This means for user space write, we don't really have an equivalent of
> --next-fua to find super block write.
>
> So this patch will add a new option, --next-write <sector> to
> replay/find next write to certain sector.
> And normally the <sector> should be super block sector number.
>
> With that, we can replay to super block write even it's user space
> triggering the write.
>
> Signed-off-by: Qu Wenruo <wqu@suse.com>
This lools fine to me. But I'd like Josef or Amir to take a look at it
too, I think they're more familiar with log-writes infrastructure than
me :)
Thanks,
Eryu
> ---
> src/log-writes/replay-log.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/src/log-writes/replay-log.c b/src/log-writes/replay-log.c
> index c16df40e5741..0d0ec49c8a00 100644
> --- a/src/log-writes/replay-log.c
> +++ b/src/log-writes/replay-log.c
> @@ -9,6 +9,7 @@
> enum option_indexes {
> NEXT_FLUSH,
> NEXT_FUA,
> + NEXT_WRITE,
> START_ENTRY,
> END_MARK,
> LOG,
> @@ -28,6 +29,7 @@ enum option_indexes {
> static struct option long_options[] = {
> {"next-flush", no_argument, NULL, 0},
> {"next-fua", no_argument, NULL, 0},
> + {"next-write", required_argument, NULL, 0},
> {"start-entry", required_argument, NULL, 0},
> {"end-mark", required_argument, NULL, 0},
> {"log", required_argument, NULL, 0},
> @@ -53,6 +55,8 @@ static void usage(void)
> fprintf(stderr, "\t--limit <number> - number of entries to replay\n");
> fprintf(stderr, "\t--next-flush - replay to/find the next flush\n");
> fprintf(stderr, "\t--next-fua - replay to/find the next fua\n");
> + fprintf(stderr, "\t--next-write <sector> - replay to/find the next "
> + "write to <sector>\n");
> fprintf(stderr, "\t--start-entry <entry> - start at the given "
> "entry #\n");
> fprintf(stderr, "\t--start-mark <mark> - mark to start from\n");
> @@ -139,6 +143,7 @@ int main(int argc, char **argv)
> char *logfile = NULL, *replayfile = NULL, *fsck_command = NULL;
> struct log_write_entry *entry;
> u64 stop_flags = 0;
> + u64 stop_sector = 0;
> u64 start_entry = 0;
> u64 start_sector = 0;
> u64 end_sector = -1ULL;
> @@ -173,6 +178,14 @@ int main(int argc, char **argv)
> case NEXT_FUA:
> stop_flags |= LOG_FUA_FLAG;
> break;
> + case NEXT_WRITE:
> + stop_sector = strtoull(optarg, &tmp, 0);
> + if (tmp && *tmp != '\0') {
> + fprintf(stderr, "Invalid sector number\n");
> + exit(1);
> + }
> + tmp = NULL;
> + break;
> case START_ENTRY:
> start_entry = strtoull(optarg, &tmp, 0);
> if (tmp && *tmp != '\0') {
> @@ -324,7 +337,8 @@ int main(int argc, char **argv)
> while ((ret = log_seek_next_entry(log, entry, 1)) == 0) {
> num_entries++;
> if ((run_limit && num_entries == run_limit) ||
> - should_stop(entry, stop_flags, end_mark)) {
> + should_stop(entry, stop_flags, end_mark) ||
> + (stop_sector && entry->sector == stop_sector)) {
> printf("%llu@%llu\n",
> (unsigned long long)log->cur_entry - 1,
> log->cur_pos / log->sectorsize);
> --
> 2.21.0
>
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] fstests: log-writes: Add new option to replay/find next write to sector
2019-03-28 5:35 [PATCH] fstests: log-writes: Add new option to replay/find next write to sector Qu Wenruo
2019-04-06 12:10 ` Eryu Guan
@ 2019-04-06 16:49 ` Amir Goldstein
1 sibling, 0 replies; 3+ messages in thread
From: Amir Goldstein @ 2019-04-06 16:49 UTC (permalink / raw)
To: Qu Wenruo; +Cc: Linux Btrfs, fstests
On Thu, Mar 28, 2019 at 7:36 AM Qu Wenruo <wqu@suse.com> wrote:
>
> For kernel operation we have METADATA/FUA/FLUSH flags as a beacon, but
> for user space write we don't have any useful flag unless the user space
> tool call fsync() to generate a FLUSH bio.
>
> This means for user space write, we don't really have an equivalent of
> --next-fua to find super block write.
>
> So this patch will add a new option, --next-write <sector> to
> replay/find next write to certain sector.
> And normally the <sector> should be super block sector number.
>
> With that, we can replay to super block write even it's user space
> triggering the write.
>
> Signed-off-by: Qu Wenruo <wqu@suse.com>
This is useful, but it could be simpler and even more useful.
See suggestion below.
> ---
> src/log-writes/replay-log.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/src/log-writes/replay-log.c b/src/log-writes/replay-log.c
> index c16df40e5741..0d0ec49c8a00 100644
> --- a/src/log-writes/replay-log.c
> +++ b/src/log-writes/replay-log.c
> @@ -9,6 +9,7 @@
> enum option_indexes {
> NEXT_FLUSH,
> NEXT_FUA,
> + NEXT_WRITE,
> START_ENTRY,
> END_MARK,
> LOG,
> @@ -28,6 +29,7 @@ enum option_indexes {
> static struct option long_options[] = {
> {"next-flush", no_argument, NULL, 0},
> {"next-fua", no_argument, NULL, 0},
> + {"next-write", required_argument, NULL, 0},
Suggest no_argument like next-fua,next-flush and assign
a special flag or something to find WRITE entries
because start/end-sector already gives to the ability
to filter by single sector or range of sectors.
If fact I added these option to debug the exact case you are
describing.
> {"start-entry", required_argument, NULL, 0},
> {"end-mark", required_argument, NULL, 0},
> {"log", required_argument, NULL, 0},
> @@ -53,6 +55,8 @@ static void usage(void)
> fprintf(stderr, "\t--limit <number> - number of entries to replay\n");
> fprintf(stderr, "\t--next-flush - replay to/find the next flush\n");
> fprintf(stderr, "\t--next-fua - replay to/find the next fua\n");
> + fprintf(stderr, "\t--next-write <sector> - replay to/find the next "
> + "write to <sector>\n");
> fprintf(stderr, "\t--start-entry <entry> - start at the given "
> "entry #\n");
> fprintf(stderr, "\t--start-mark <mark> - mark to start from\n");
> @@ -139,6 +143,7 @@ int main(int argc, char **argv)
> char *logfile = NULL, *replayfile = NULL, *fsck_command = NULL;
> struct log_write_entry *entry;
> u64 stop_flags = 0;
> + u64 stop_sector = 0;
> u64 start_entry = 0;
> u64 start_sector = 0;
> u64 end_sector = -1ULL;
> @@ -173,6 +178,14 @@ int main(int argc, char **argv)
> case NEXT_FUA:
> stop_flags |= LOG_FUA_FLAG;
> break;
> + case NEXT_WRITE:
> + stop_sector = strtoull(optarg, &tmp, 0);
> + if (tmp && *tmp != '\0') {
> + fprintf(stderr, "Invalid sector number\n");
> + exit(1);
> + }
> + tmp = NULL;
> + break;
> case START_ENTRY:
> start_entry = strtoull(optarg, &tmp, 0);
> if (tmp && *tmp != '\0') {
> @@ -324,7 +337,8 @@ int main(int argc, char **argv)
> while ((ret = log_seek_next_entry(log, entry, 1)) == 0) {
> num_entries++;
> if ((run_limit && num_entries == run_limit) ||
> - should_stop(entry, stop_flags, end_mark)) {
> + should_stop(entry, stop_flags, end_mark) ||
> + (stop_sector && entry->sector == stop_sector)) {
You seem to have only implemented "find" mode and not "replay" mode
(the log_replay_next_entry() loop).
I suggest that you work log_should_skip() into both "find" and "replay"
loops. i.e. (!log_should_skip() && should_stop()).
Thanks,
Amir.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-04-06 16:49 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-03-28 5:35 [PATCH] fstests: log-writes: Add new option to replay/find next write to sector Qu Wenruo
2019-04-06 12:10 ` Eryu Guan
2019-04-06 16:49 ` Amir Goldstein
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).