| Message ID | 5607133.DvuYhMxLoT@kreacher |
|---|---|
| State | New |
| Headers | show |
| Series | [v2] ata: ahci: Do not check ACPI_FADT_LOW_POWER_S0 | expand |
On 8/25/2022 13:01, Rafael J. Wysocki wrote: > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > The ACPI_FADT_LOW_POWER_S0 flag merely means that it is better to > use low-power S0 idle on the given platform than S3 (provided that > the latter is supported) and it doesn't preclude using either of > them (which of them will be used depends on the choices made by user > space). > > For this reason, there is no benefit from checking that flag in > ahci_update_initial_lpm_policy(). > > First off, it cannot be a bug to do S3 with policy set to either > ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER, because S3 can be > used on systems with ACPI_FADT_LOW_POWER_S0 set and it must work if > really supported, so the ACPI_FADT_LOW_POWER_S0 check is not needed to > protect the S3-capable systems from failing. > > Second, suspend-to-idle can be carried out on a system with > ACPI_FADT_LOW_POWER_S0 unset and it is expected to work, so if setting > policy to either ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER is > needed to handle that case correctly, it should be done regardless of > the ACPI_FADT_LOW_POWER_S0 value. > > Accordingly, drop the ACPI_FADT_LOW_POWER_S0 check from > ahci_update_initial_lpm_policy() along with the CONFIG_ACPI #ifdef > around it that is not necessary any more. Looking at the source commit for this behavior: b1a9585cc396 ("ata: ahci: Enable DEVSLP by default on x86 with SLP_S0") It was trying to set a policy tied to when the system is defaulting to suspend to idle. To try to match the spirit of the original request but not tying it to the FADT, how about using pm_suspend_default_s2idle()? > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > --- > > v1 -> v2: > * Adjust subject (Damien). > * Drop #ifdef CONFIG_ACPI that is not necessary any more (Mario). > * Update the changelog. > > --- > drivers/ata/ahci.c | 5 +---- > 1 file changed, 1 insertion(+), 4 deletions(-) > > Index: linux-pm/drivers/ata/ahci.c > =================================================================== > --- linux-pm.orig/drivers/ata/ahci.c > +++ linux-pm/drivers/ata/ahci.c > @@ -1609,15 +1609,12 @@ static void ahci_update_initial_lpm_poli > goto update_policy; > } > > -#ifdef CONFIG_ACPI > - if (policy > ATA_LPM_MED_POWER && > - (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) { > + if (policy > ATA_LPM_MED_POWER) { > if (hpriv->cap & HOST_CAP_PART) > policy = ATA_LPM_MIN_POWER_WITH_PARTIAL; > else if (hpriv->cap & HOST_CAP_SSC) > policy = ATA_LPM_MIN_POWER; > } > -#endif > > update_policy: > if (policy >= ATA_LPM_UNKNOWN && policy <= ATA_LPM_MIN_POWER) > > >
On Thu, Aug 25, 2022 at 8:17 PM Limonciello, Mario <mario.limonciello@amd.com> wrote: > > On 8/25/2022 13:01, Rafael J. Wysocki wrote: > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > The ACPI_FADT_LOW_POWER_S0 flag merely means that it is better to > > use low-power S0 idle on the given platform than S3 (provided that > > the latter is supported) and it doesn't preclude using either of > > them (which of them will be used depends on the choices made by user > > space). > > > > For this reason, there is no benefit from checking that flag in > > ahci_update_initial_lpm_policy(). > > > > First off, it cannot be a bug to do S3 with policy set to either > > ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER, because S3 can be > > used on systems with ACPI_FADT_LOW_POWER_S0 set and it must work if > > really supported, so the ACPI_FADT_LOW_POWER_S0 check is not needed to > > protect the S3-capable systems from failing. > > > > Second, suspend-to-idle can be carried out on a system with > > ACPI_FADT_LOW_POWER_S0 unset and it is expected to work, so if setting > > policy to either ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER is > > needed to handle that case correctly, it should be done regardless of > > the ACPI_FADT_LOW_POWER_S0 value. > > > > Accordingly, drop the ACPI_FADT_LOW_POWER_S0 check from > > ahci_update_initial_lpm_policy() along with the CONFIG_ACPI #ifdef > > around it that is not necessary any more. > > Looking at the source commit for this behavior: > > b1a9585cc396 ("ata: ahci: Enable DEVSLP by default on x86 with SLP_S0") > > It was trying to set a policy tied to when the system is defaulting to > suspend to idle. > > To try to match the spirit of the original request but not tying it to > the FADT, how about using pm_suspend_default_s2idle()? The user can switch to "default S3" later anyway, so this wouldn't help more than the check being dropped.
On Thu, Aug 25, 2022 at 8:29 PM Limonciello, Mario <mario.limonciello@amd.com> wrote: > > On 8/25/2022 13:26, Rafael J. Wysocki wrote: > > On Thu, Aug 25, 2022 at 8:17 PM Limonciello, Mario > > <mario.limonciello@amd.com> wrote: > >> > >> On 8/25/2022 13:01, Rafael J. Wysocki wrote: > >>> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > >>> > >>> The ACPI_FADT_LOW_POWER_S0 flag merely means that it is better to > >>> use low-power S0 idle on the given platform than S3 (provided that > >>> the latter is supported) and it doesn't preclude using either of > >>> them (which of them will be used depends on the choices made by user > >>> space). > >>> > >>> For this reason, there is no benefit from checking that flag in > >>> ahci_update_initial_lpm_policy(). > >>> > >>> First off, it cannot be a bug to do S3 with policy set to either > >>> ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER, because S3 can be > >>> used on systems with ACPI_FADT_LOW_POWER_S0 set and it must work if > >>> really supported, so the ACPI_FADT_LOW_POWER_S0 check is not needed to > >>> protect the S3-capable systems from failing. > >>> > >>> Second, suspend-to-idle can be carried out on a system with > >>> ACPI_FADT_LOW_POWER_S0 unset and it is expected to work, so if setting > >>> policy to either ATA_LPM_MIN_POWER_WITH_PARTIAL or ATA_LPM_MIN_POWER is > >>> needed to handle that case correctly, it should be done regardless of > >>> the ACPI_FADT_LOW_POWER_S0 value. > >>> > >>> Accordingly, drop the ACPI_FADT_LOW_POWER_S0 check from > >>> ahci_update_initial_lpm_policy() along with the CONFIG_ACPI #ifdef > >>> around it that is not necessary any more. > >> > >> Looking at the source commit for this behavior: > >> > >> b1a9585cc396 ("ata: ahci: Enable DEVSLP by default on x86 with SLP_S0") > >> > >> It was trying to set a policy tied to when the system is defaulting to > >> suspend to idle. > >> > >> To try to match the spirit of the original request but not tying it to > >> the FADT, how about using pm_suspend_default_s2idle()? > > > > The user can switch to "default S3" later anyway, so this wouldn't > > help more than the check being dropped. > > Right, they could also change LPM policy to different policy later too > if they want. Exactly. > This is just for setting up default policy. I think if you matched to > only when pm_suspend_default_s2idle() it would be the least likelihood > to change this default policy on unsuspecting people upgrading. The only case where it matters is systems doing S3 by default that would end up enabling DEVSLP. Would that confuse the BIOSes on them? Maybe, but I think that S3 with DEVSLP enabled is generally expected to work. Anyway, I'm not religious about this, so I'll send a v3.
Index: linux-pm/drivers/ata/ahci.c =================================================================== --- linux-pm.orig/drivers/ata/ahci.c +++ linux-pm/drivers/ata/ahci.c @@ -1609,15 +1609,12 @@ static void ahci_update_initial_lpm_poli goto update_policy; } -#ifdef CONFIG_ACPI - if (policy > ATA_LPM_MED_POWER && - (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) { + if (policy > ATA_LPM_MED_POWER) { if (hpriv->cap & HOST_CAP_PART) policy = ATA_LPM_MIN_POWER_WITH_PARTIAL; else if (hpriv->cap & HOST_CAP_SSC) policy = ATA_LPM_MIN_POWER; } -#endif update_policy: if (policy >= ATA_LPM_UNKNOWN && policy <= ATA_LPM_MIN_POWER)