Message ID | 20191030174401.10926-1-alex.hung@canonical.com |
---|---|
State | Accepted |
Headers | show |
Series | acpi/s3: fine-tune measure of suspend time | expand |
On 30/10/2019 18:44, Alex Hung wrote: > The time variation can be reduced. For example, the original suspend > time: > Suspend: 3.966 seconds. > Suspend: 0.539 seconds. > Suspend: 1.156 seconds. > After patched: > Suspend: 0.486 seconds. > Suspend: 0.508 seconds. > Suspend: 0.528 seconds. > > This is especially true for 1st suspend when kernel is busy with other > kernel modules such as > > [ 120.123903] fwts: Suspend > [ 120.162215] wlp60s0: deauthenticating from ... > [ 123.726367] PM: suspend entry (deep) > > Signed-off-by: Alex Hung <alex.hung@canonical.com> Thanks Alex. Acked-by: Colin Ian King <colin.king@canonical.com> > --- > src/acpi/s3/s3.c | 7 +++++++ > 1 file changed, 7 insertions(+) > > diff --git a/src/acpi/s3/s3.c b/src/acpi/s3/s3.c > index 6664db2c..5ec8d5c8 100644 > --- a/src/acpi/s3/s3.c > +++ b/src/acpi/s3/s3.c > @@ -411,6 +411,13 @@ static int s3_scan_times( > s3_suspend_start = ts; > continue; > } > + > + /* Update log time if this is available */ > + if (strstr(txt, "PM: suspend entry")) { > + s3_suspend_start = ts; > + continue; > + } > + > if (strstr(txt, FWTS_RESUME)) { > s3_resume_finish = ts; > continue; >
On Thu, Oct 31, 2019 at 1:44 AM Alex Hung <alex.hung@canonical.com> wrote: > > The time variation can be reduced. For example, the original suspend > time: > Suspend: 3.966 seconds. > Suspend: 0.539 seconds. > Suspend: 1.156 seconds. > After patched: > Suspend: 0.486 seconds. > Suspend: 0.508 seconds. > Suspend: 0.528 seconds. > > This is especially true for 1st suspend when kernel is busy with other > kernel modules such as > > [ 120.123903] fwts: Suspend > [ 120.162215] wlp60s0: deauthenticating from ... > [ 123.726367] PM: suspend entry (deep) > > Signed-off-by: Alex Hung <alex.hung@canonical.com> > --- > src/acpi/s3/s3.c | 7 +++++++ > 1 file changed, 7 insertions(+) > > diff --git a/src/acpi/s3/s3.c b/src/acpi/s3/s3.c > index 6664db2c..5ec8d5c8 100644 > --- a/src/acpi/s3/s3.c > +++ b/src/acpi/s3/s3.c > @@ -411,6 +411,13 @@ static int s3_scan_times( > s3_suspend_start = ts; > continue; > } > + > + /* Update log time if this is available */ > + if (strstr(txt, "PM: suspend entry")) { > + s3_suspend_start = ts; > + continue; > + } > + > if (strstr(txt, FWTS_RESUME)) { > s3_resume_finish = ts; > continue; > -- > 2.17.1 Thanks Alex for looking at this issue. Seems if we can also record the time between fwts starts suspending and pm_suspend() is called, that's still valuable for discovering issues that takes a long time before the system can go to suspend. Acked-by: Anthony Wong <anthony.wong@canonical.com>
diff --git a/src/acpi/s3/s3.c b/src/acpi/s3/s3.c index 6664db2c..5ec8d5c8 100644 --- a/src/acpi/s3/s3.c +++ b/src/acpi/s3/s3.c @@ -411,6 +411,13 @@ static int s3_scan_times( s3_suspend_start = ts; continue; } + + /* Update log time if this is available */ + if (strstr(txt, "PM: suspend entry")) { + s3_suspend_start = ts; + continue; + } + if (strstr(txt, FWTS_RESUME)) { s3_resume_finish = ts; continue;
The time variation can be reduced. For example, the original suspend time: Suspend: 3.966 seconds. Suspend: 0.539 seconds. Suspend: 1.156 seconds. After patched: Suspend: 0.486 seconds. Suspend: 0.508 seconds. Suspend: 0.528 seconds. This is especially true for 1st suspend when kernel is busy with other kernel modules such as [ 120.123903] fwts: Suspend [ 120.162215] wlp60s0: deauthenticating from ... [ 123.726367] PM: suspend entry (deep) Signed-off-by: Alex Hung <alex.hung@canonical.com> --- src/acpi/s3/s3.c | 7 +++++++ 1 file changed, 7 insertions(+)