This patch adds more debug logging regarding to plock operation time.
Currently plocks are handled by user space and we are waiting for a
reply of the user space. To measure this upcall and downcall again we
introduce logging information of how much time the response took and
additional posix lock
This patch adds more debug logging regarding to plock operation time.
Currently plocks are handled by user space and we are waiting for a
reply of the user space. To measure this upcall and downcall again we
introduce logging information of how much time the response took and
additional posix lock
Hi,
On Mon, Jan 30, 2023 at 2:24 PM Alexander Aring wrote:
>
> This patch will trace lock state changes of the used dlm posix locks. In
> combination with the plock logfile we can see state changes over time
> and follow posix locks and their state.
> ---
> dlm_controld/dlm_daemon.h | 2 +-
>
Hi,
On Mon, Jan 30, 2023 at 2:24 PM Alexander Aring wrote:
>
> This patch adds functionality to see how long a posix lock is alive or
> is in waiting or pending state. It can be used to filter out interesting
> locks which are stuck in e.g. waiting state to know that a user space
> process probab
This patch adds functionality to see how long a posix lock is alive or
is in waiting or pending state. It can be used to filter out interesting
locks which are stuck in e.g. waiting state to know that a user space
process probably has contention on it. The logging information will
printout addition
The current plock logging is limited due a in-memory log buffer which
can be dumped via dlm_contol log_plock functionality. To trace plock
performance issues it's necessary to log plock activity in a bigger log
buffer such as a file. This patch will add functionality that plock
logging information
Currently dlm_controld sets a embedded data int value of dlm_controld
dump functionality failed for e.g. if lockspace cannot be found. The
dlm_tool does not parse this possible error functionality and will exit
successfully. This patch will add dlm_tool fail functionality if
dlm_controld sets an em
Those parameters are only used read only. We don't change any data where
those pointers point to.
---
dlm_controld/plock.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/dlm_controld/plock.c b/dlm_controld/plock.c
index 77c043fd..d83a79d2 100644
--- a/dlm_controld/plock.c
+++
This patch will trace lock state changes of the used dlm posix locks. In
combination with the plock logfile we can see state changes over time
and follow posix locks and their state.
---
dlm_controld/dlm_daemon.h | 2 +-
dlm_controld/logging.c| 2 +-
dlm_controld/plock.c | 124 +++
The ls parameter in write_result() is not used, so we can remove it.
---
dlm_controld/plock.c | 13 ++---
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/dlm_controld/plock.c b/dlm_controld/plock.c
index c2c80360..462c9212 100644
--- a/dlm_controld/plock.c
+++ b/dlm_controld/
Currently the logdir will be created only if logfile does contain a
string. To add another logfiles we simple create the logdir always on
startup.
---
dlm_controld/logging.c | 34 +-
1 file changed, 17 insertions(+), 17 deletions(-)
diff --git a/dlm_controld/loggin
Ad the loglevel is for save plock messages during corosync resource
membership upate. This patch will put the processing of saved messages
on the same loglevel.
---
dlm_controld/plock.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/dlm_controld/plock.c b/dlm_controld/ploc
This set is mainly cleanups but patches 5 and 6 are small fixes for
fsck.gfs2 which I would like to draw attention to.
PR is https://pagure.io/gfs2-utils/pull-request/14 awaiting CI testing
should you prefer to comment there.
Andy
Andrew Price (6):
libgfs2: Return the inode from lgfs2_lookupi(
It always points to 'dentry' so it can be replaced in all uses.
Signed-off-by: Andrew Price
---
gfs2/fsck/pass2.c | 13 ++---
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/gfs2/fsck/pass2.c b/gfs2/fsck/pass2.c
index 0c2e0146..002bf2cb 100644
--- a/gfs2/fsck/pass2.c
+++ b/
fsck.gfs2 can leave i_blocks too large when it removes empty leaf
blocks, so decrease the count when freeing them.
Signed-off-by: Andrew Price
---
gfs2/fsck/pass2.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/gfs2/fsck/pass2.c b/gfs2/fsck/pass2.c
index 5327ee43..71459978 100644
--- a/gfs
'dent' points to the on-disk data that 'dentry' was converted from, so
this comparison always evaluates to true. Compare to 'entry' instead.
Signed-off-by: Andrew Price
---
gfs2/fsck/pass2.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gfs2/fsck/pass2.c b/gfs2/fsck/pass2.c
Instead of returning a -errno with the inode set via a passed in
pointer, return the inode and use errno. This is more consistent and
simplifies some calling code.
Signed-off-by: Andrew Price
---
gfs2/convert/gfs2_convert.c | 7 +++
gfs2/edit/hexedit.c | 2 +-
gfs2/fsck/fs_recovery
Move the lookup for an existing inode to lgfs2_createi itself and then
create the inode in __createi (renamed to do_createi) unconditionally.
Signed-off-by: Andrew Price
---
gfs2/libgfs2/fs_ops.c | 80 ---
1 file changed, 45 insertions(+), 35 deletions(-)
The one caller sets sdp->gfs1 and modifies the mode itself so we can use
sdp->gfs1 in __createi() and update the caller to use lgfs2_createi()
instead.
Signed-off-by: Andrew Price
---
gfs2/fsck/lost_n_found.c | 6 +-
gfs2/libgfs2/fs_ops.c| 19 ++-
gfs2/libgfs2/libgfs2.h
Hello Edward,
On Fri, Jan 27, 2023 at 6:12 AM wrote:
> From: Edward Adam Davis
>
> [ 81.372851][ T5532] CPU: 1 PID: 5532 Comm: syz-executor.0 Not tainted
> 6.2.0-rc1-syzkaller-dirty #0
> [ 81.382080][ T5532] Hardware name: Google Google Compute Engine/Google
> Compute Engine, BIOS Google 0
From: Edward Adam Davis
[ 81.372851][ T5532] CPU: 1 PID: 5532 Comm: syz-executor.0 Not tainted
6.2.0-rc1-syzkaller-dirty #0
[ 81.382080][ T5532] Hardware name: Google Google Compute Engine/Google
Compute Engine, BIOS Google 01/12/2023
[ 81.392343][ T5532] Call Trace:
[ 81.395654][ T5532
21 matches
Mail list logo
