Re: [REGRESSION] 2.6.29=>2.6.30 suspend to ram regression (IDE related)

From: Bartlomiej Zolnierkiewicz
Date: Sun Jun 28 2009 - 09:21:00 EST


On Sunday 28 June 2009 10:06:02 Etienne Basset wrote:
> Jeff Chua wrote:
> > On Sun, Jun 28, 2009 at 2:09 AM, Jeff Chua<jeff.chua.linux@xxxxxxxxx> wrote:
> >> On Sat, Jun 27, 2009 at 5:04 PM, Etienne
> >> Basset<etienne.basset@xxxxxxxxxxxxxx> wrote:
> >>>>> kernel v2.6.29 suspends to RAM reliably on my computer; v2.6.30 doesn't resume after STR
> >> Same problem on Thinkpad X61. STD STR suspended ok. Upon resume,
> >> everything "seems" fine for about 100 seconds. Then mouse, keyboard
> >> frozen.
> >>
> >>>>> I tried also 2.6.31-rc1 doesn't work either
> >>>>> etienne@etienne-desktop:~/linux-2.6$ git bisect bad
> >>>>> 2f0d0fd2a605666d38e290c5c0d2907484352dc4 is first bad commit
> >>>> Does the following patch fix it?
> >>> Yes, it works now
> >> Works for me too with the patch.

Thanks for testing (also sorry for the problem, hopefully the fix will
prevent any similar issues in the future).

> > UPDATE ...
> >
> > STR able to resume and works after >100 seconds
> >
> > STD able to resume but failed after 100 seconds (could be 300 seconds)
> > but eventually failed. Everything seems ok, but mouse and keyboard
> > just freezed suddenly. Just haven't really time it enough.
> >
> > Jeff.
> >
> Hello,
>
> same here ; after resume, computer hangs after 2minutes (cannot ping from outside)
> and STR/resume doesn't work if I try from console not from X
> there must be at least one another bug lurking
> If someone hasn't a better idea I'll try another bisection, applying bart's patch at each step

I think that you're right. It seems that the underlying issue could be
a non-working/broken ACPI support (please note that the IDE ACPI bug won't
be triggered otherwise).

David, please apply:

From: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
Subject: [PATCH] ide: fix resume for CONFIG_BLK_DEV_IDEACPI=y

commit 2f0d0fd2a605666d38e290c5c0d2907484352dc4 ("ide-acpi: cleanup
do_drive_get_GTF()") didn't account for the lack of hwif->acpidata
check in generic_ide_suspend() [ indirect user of do_drive_get_GTF()
through ide_acpi_exec_tfs() ] resulting in broken resume when ACPI
support is enabled but ACPI data is unavailable.

Fix it by adding ide_port_acpi() helper for checking if port needs
ACPI handling and cleaning generic_ide_{suspend,resume}() to use it
instead of hiding hwif->acpidata and ide_noacpi checks in IDE ACPI
helpers (this should help in preventing similar bugs in the future).

While at it:
- kill superfluous debugging printks in ide_acpi_{get,push}_timing()

Reported-and-tested-by: Etienne Basset <etienne.basset@xxxxxxxxxxxxxx>
Also-reported-and-tested-by: Jeff Chua <jeff.chua.linux@xxxxxxxxx>
Cc: "Rafael J. Wysocki" <rjw@xxxxxxx>
Cc: stable@xxxxxxxxxx
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
---
Added patch description, the patch itself remains unchanged.

drivers/ide/ide-acpi.c | 37 +++++++------------------------------
drivers/ide/ide-pm.c | 30 ++++++++++++++++++------------
include/linux/ide.h | 2 ++
3 files changed, 27 insertions(+), 42 deletions(-)

Index: b/drivers/ide/ide-acpi.c
===================================================================
--- a/drivers/ide/ide-acpi.c
+++ b/drivers/ide/ide-acpi.c
@@ -92,6 +92,11 @@ int ide_acpi_init(void)
return 0;
}

+bool ide_port_acpi(ide_hwif_t *hwif)
+{
+ return ide_noacpi == 0 && hwif->acpidata;
+}
+
/**
* ide_get_dev_handle - finds acpi_handle and PCI device.function
* @dev: device to locate
@@ -352,9 +357,6 @@ int ide_acpi_exec_tfs(ide_drive_t *drive
unsigned long gtf_address;
unsigned long obj_loc;

- if (ide_noacpi)
- return 0;
-
DEBPRINT("call get_GTF, drive=%s port=%d\n", drive->name, drive->dn);

ret = do_drive_get_GTF(drive, &gtf_length, &gtf_address, &obj_loc);
@@ -389,16 +391,6 @@ void ide_acpi_get_timing(ide_hwif_t *hwi
struct acpi_buffer output;
union acpi_object *out_obj;

- if (ide_noacpi)
- return;
-
- DEBPRINT("ENTER:\n");
-
- if (!hwif->acpidata) {
- DEBPRINT("no ACPI data for %s\n", hwif->name);
- return;
- }
-
/* Setting up output buffer for _GTM */
output.length = ACPI_ALLOCATE_BUFFER;
output.pointer = NULL; /* ACPI-CA sets this; save/free it later */
@@ -479,16 +471,6 @@ void ide_acpi_push_timing(ide_hwif_t *hw
struct ide_acpi_drive_link *master = &hwif->acpidata->master;
struct ide_acpi_drive_link *slave = &hwif->acpidata->slave;

- if (ide_noacpi)
- return;
-
- DEBPRINT("ENTER:\n");
-
- if (!hwif->acpidata) {
- DEBPRINT("no ACPI data for %s\n", hwif->name);
- return;
- }
-
/* Give the GTM buffer + drive Identify data to the channel via the
* _STM method: */
/* setup input parameters buffer for _STM */
@@ -527,16 +509,11 @@ void ide_acpi_set_state(ide_hwif_t *hwif
ide_drive_t *drive;
int i;

- if (ide_noacpi || ide_noacpi_psx)
+ if (ide_noacpi_psx)
return;

DEBPRINT("ENTER:\n");

- if (!hwif->acpidata) {
- DEBPRINT("no ACPI data for %s\n", hwif->name);
- return;
- }
-
/* channel first and then drives for power on and verse versa for power off */
if (on)
acpi_bus_set_power(hwif->acpidata->obj_handle, ACPI_STATE_D0);
@@ -616,7 +593,7 @@ void ide_acpi_port_init_devices(ide_hwif
drive->name, err);
}

- if (!ide_acpionboot) {
+ if (ide_noacpi || ide_acpionboot == 0) {
DEBPRINT("ACPI methods disabled on boot\n");
return;
}
Index: b/drivers/ide/ide-pm.c
===================================================================
--- a/drivers/ide/ide-pm.c
+++ b/drivers/ide/ide-pm.c
@@ -10,9 +10,11 @@ int generic_ide_suspend(struct device *d
struct request_pm_state rqpm;
int ret;

- /* call ACPI _GTM only once */
- if ((drive->dn & 1) == 0 || pair == NULL)
- ide_acpi_get_timing(hwif);
+ if (ide_port_acpi(hwif)) {
+ /* call ACPI _GTM only once */
+ if ((drive->dn & 1) == 0 || pair == NULL)
+ ide_acpi_get_timing(hwif);
+ }

memset(&rqpm, 0, sizeof(rqpm));
rq = blk_get_request(drive->queue, READ, __GFP_WAIT);
@@ -26,9 +28,11 @@ int generic_ide_suspend(struct device *d
ret = blk_execute_rq(drive->queue, NULL, rq, 0);
blk_put_request(rq);

- /* call ACPI _PS3 only after both devices are suspended */
- if (ret == 0 && ((drive->dn & 1) || pair == NULL))
- ide_acpi_set_state(hwif, 0);
+ if (ret == 0 && ide_port_acpi(hwif)) {
+ /* call ACPI _PS3 only after both devices are suspended */
+ if ((drive->dn & 1) || pair == NULL)
+ ide_acpi_set_state(hwif, 0);
+ }

return ret;
}
@@ -42,13 +46,15 @@ int generic_ide_resume(struct device *de
struct request_pm_state rqpm;
int err;

- /* call ACPI _PS0 / _STM only once */
- if ((drive->dn & 1) == 0 || pair == NULL) {
- ide_acpi_set_state(hwif, 1);
- ide_acpi_push_timing(hwif);
- }
+ if (ide_port_acpi(hwif)) {
+ /* call ACPI _PS0 / _STM only once */
+ if ((drive->dn & 1) == 0 || pair == NULL) {
+ ide_acpi_set_state(hwif, 1);
+ ide_acpi_push_timing(hwif);
+ }

- ide_acpi_exec_tfs(drive);
+ ide_acpi_exec_tfs(drive);
+ }

memset(&rqpm, 0, sizeof(rqpm));
rq = blk_get_request(drive->queue, READ, __GFP_WAIT);
Index: b/include/linux/ide.h
===================================================================
--- a/include/linux/ide.h
+++ b/include/linux/ide.h
@@ -1420,6 +1420,7 @@ static inline void ide_dma_unmap_sg(ide_

#ifdef CONFIG_BLK_DEV_IDEACPI
int ide_acpi_init(void);
+bool ide_port_acpi(ide_hwif_t *hwif);
extern int ide_acpi_exec_tfs(ide_drive_t *drive);
extern void ide_acpi_get_timing(ide_hwif_t *hwif);
extern void ide_acpi_push_timing(ide_hwif_t *hwif);
@@ -1428,6 +1429,7 @@ void ide_acpi_port_init_devices(ide_hwif
extern void ide_acpi_set_state(ide_hwif_t *hwif, int on);
#else
static inline int ide_acpi_init(void) { return 0; }
+static inline bool ide_port_acpi(ide_hwif_t *hwif) { return 0; }
static inline int ide_acpi_exec_tfs(ide_drive_t *drive) { return 0; }
static inline void ide_acpi_get_timing(ide_hwif_t *hwif) { ; }
static inline void ide_acpi_push_timing(ide_hwif_t *hwif) { ; }
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/