[PATCH v2] nvme: explicitly disable APST on quirked devices

[PATCH v2] nvme: explicitly disable APST on quirked devices

[Kai-Heng Feng]

A user reports APST is enabled, even when the NVMe is quirked or with
option "default_ps_max_latency_us=0".

The current logic will not set APST if the device is quirked. But the
NVMe in question will enable APST automatically.

Separate the logic "apst is supported" and "to enable apst", so we can
use the latter one to explicitly disable APST at initialiaztion.

BugLink: https://bugs.launchpad.net/bugs/1699004
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
---
v2:
 - Remove unnecessary check on device state.
 - Correct apst_enabled value when there's no APST quirk matched.

 drivers/nvme/host/core.c | 17 +++++++++--------
 drivers/nvme/host/nvme.h |  1 +
 2 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 903d5813023a..f0fd1bba3cdb 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -1372,7 +1372,7 @@ static void nvme_configure_apst(struct nvme_ctrl *ctrl)
 	if (!table)
 		return;
 
-	if (ctrl->ps_max_latency_us == 0) {
+	if (!ctrl->apst_enabled || ctrl->ps_max_latency_us == 0) {
 		/* Turn off APST. */
 		apste = 0;
 		dev_dbg(ctrl->device, "APST disabled\n");
@@ -1539,7 +1539,7 @@ int nvme_init_identify(struct nvme_ctrl *ctrl)
 	u64 cap;
 	int ret, page_shift;
 	u32 max_hw_sectors;
-	u8 prev_apsta;
+	bool prev_apst_enabled;
 
 	ret = ctrl->ops->reg_read32(ctrl, NVME_REG_VS, &ctrl->vs);
 	if (ret) {
@@ -1607,16 +1607,17 @@ int nvme_init_identify(struct nvme_ctrl *ctrl)
 	ctrl->kas = le16_to_cpu(id->kas);
 
 	ctrl->npss = id->npss;
-	prev_apsta = ctrl->apsta;
+	ctrl->apsta = id->apsta;
+	prev_apst_enabled = ctrl->apst_enabled;
 	if (ctrl->quirks & NVME_QUIRK_NO_APST) {
 		if (force_apst && id->apsta) {
 			dev_warn(ctrl->dev, "forcibly allowing APST due to nvme_core.force_apst -- use at your own risk\n");
-			ctrl->apsta = 1;
+			ctrl->apst_enabled = true;
 		} else {
-			ctrl->apsta = 0;
+			ctrl->apst_enabled = false;
 		}
 	} else {
-		ctrl->apsta = id->apsta;
+		ctrl->apst_enabled = id->apsta;
 	}
 	memcpy(ctrl->psd, id->psd, sizeof(ctrl->psd));
 
@@ -1644,9 +1645,9 @@ int nvme_init_identify(struct nvme_ctrl *ctrl)
 
 	kfree(id);
 
-	if (ctrl->apsta && !prev_apsta)
+	if (ctrl->apst_enabled && !prev_apst_enabled)
 		dev_pm_qos_expose_latency_tolerance(ctrl->device);
-	else if (!ctrl->apsta && prev_apsta)
+	else if (!ctrl->apst_enabled && prev_apst_enabled)
 		dev_pm_qos_hide_latency_tolerance(ctrl->device);
 
 	nvme_configure_apst(ctrl);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 9d6a070d4391..dbabe2b728cb 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -165,6 +165,7 @@ struct nvme_ctrl {
 
 	/* Power saving configuration */
 	u64 ps_max_latency_us;
+	bool apst_enabled;
 
 	/* Fabrics only */
 	u16 sqsize;
-- 
2.13.1

Re: [PATCH v2] nvme: explicitly disable APST on quirked devices

[Andy Lutomirski]

On Mon, Jun 26, 2017 at 12:01 AM, Kai-Heng Feng
<kai.heng.feng@canonical.com> wrote:
> A user reports APST is enabled, even when the NVMe is quirked or with
> option "default_ps_max_latency_us=0".
>
> The current logic will not set APST if the device is quirked. But the
> NVMe in question will enable APST automatically.
>
> Separate the logic "apst is supported" and "to enable apst", so we can
> use the latter one to explicitly disable APST at initialiaztion.

Reviewed-by: Andy Lutomirski <luto@kernel.org>

That being said, I smell a giant WTF here.  The affected hardware
seems to have APST on by default, and APST is buggy so the disk stops
working when APST is on.  So here's the $1M question: how does the
system *boot*?  After all, it's running for a while before the kernel
gets around to turning off APST, and I really doubt that BIOS does
this.

Here's a wild theory: what if the problem on all these disks is
actually our CSTS polling?  Could it be that some of the disks
implement CSTS reads in firmware and malfunction if CSTS is read while
in PS4?  This would be a blatant spec violation, but that's never
stopped anyone before...

--Andy

Re: [PATCH v2] nvme: explicitly disable APST on quirked devices

[Kai-Heng Feng]

On Tue, Jun 27, 2017 at 2:05 AM, Andy Lutomirski <luto@kernel.org> wrote:
> On Mon, Jun 26, 2017 at 12:01 AM, Kai-Heng Feng
> <kai.heng.feng@canonical.com> wrote:
>> A user reports APST is enabled, even when the NVMe is quirked or with
>> option "default_ps_max_latency_us=0".
>>
>> The current logic will not set APST if the device is quirked. But the
>> NVMe in question will enable APST automatically.
>>
>> Separate the logic "apst is supported" and "to enable apst", so we can
>> use the latter one to explicitly disable APST at initialiaztion.
>
> Reviewed-by: Andy Lutomirski <luto@kernel.org>
>
> That being said, I smell a giant WTF here.  The affected hardware
> seems to have APST on by default, and APST is buggy so the disk stops
> working when APST is on.  So here's the $1M question: how does the
> system *boot*?  After all, it's running for a while before the kernel
> gets around to turning off APST, and I really doubt that BIOS does
> this.

>From my experience, systems never failed to boot on those faulty
NVMes. Probably because the constantly disk read required by boot
never let the NVMe transited to PS4. The problem always occurs after
some usage after boot.

Seems like the user has a tricky system. At first, APST wasn't
enabled. It's enabled after boot with a new kernel, and it's enabled
forever. Even if it's disabled explicitly, the APST is still enabled
by default on the system. The user didn't upgrade BIOS in the interim.

>
> Here's a wild theory: what if the problem on all these disks is
> actually our CSTS polling?  Could it be that some of the disks
> implement CSTS reads in firmware and malfunction if CSTS is read while
> in PS4?  This would be a blatant spec violation, but that's never
> stopped anyone before...
>
> --Andy

Re: [PATCH v2] nvme: explicitly disable APST on quirked devices

[Keith Busch]

On Mon, Jun 26, 2017 at 12:01:29AM -0700, Kai-Heng Feng wrote:
> A user reports APST is enabled, even when the NVMe is quirked or with
> option "default_ps_max_latency_us=0".
> 
> The current logic will not set APST if the device is quirked. But the
> NVMe in question will enable APST automatically.
> 
> Separate the logic "apst is supported" and "to enable apst", so we can
> use the latter one to explicitly disable APST at initialiaztion.
> 
> BugLink: https://bugs.launchpad.net/bugs/1699004
> Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>

There's something off about the format with this patch such that it is
getting mangled when I save and apply it.

I'm not sure if the content type and encoding are the culprit, but that
just stood out as different. Normally patches I receive have this header:

  Content-Type: text/plain; charset="us-ascii"
  Content-Transfer-Encoding: 7bit

But this one has:

  Content-Type: text/plain; charset="iso-8859-1"
  Content-Transfer-Encoding: quoted-printable

In any case, I'll hand apply it with Andy's review.