Re: Hotplug borked after suspend/resume in Linux-3.3 ?

[Lin Ming <ming.m.lin@intel.com>]

On Thu, 2012-04-19 at 20:22 +0200, Martin Mokrejs wrote:
> Hi Lin,
> 
> Lin Ming wrote:
> > On Thu, Apr 19, 2012 at 2:32 AM, Martin Mokrejs <mmokrejs@fold.natur.cuni.cz> wrote:
> >> Martin Mokrejs wrote:
> >>>
> >>>
> >>> Jeff Garzik wrote:
> >>>> On 04/18/2012 01:10 PM, Martin Mokrejs wrote:
> >>>>> Fix: I got my 3TB disk detected by this single command:
> >>>>>
> >>>>> # echo on>   /sys/devices/pci0000:00/0000:00:1f.2/ata6/power/control
> >>>>> #
> >>>>>
> >>>>> This is a Dell Vostro 3550 with A09 BIOS. Same happend with 3.4-rc3 kernel.
> >>>>>
> >>>>> I can do some more testing if you want me to.
> >>>>> Best,
> >>>>> Martin
> >>>>
> >>>>
> >>>> Can you test this one-line patch from Lin Ming?  Hopefully there is zero sysfs twiddling required with this one...
> >>>>
> >>>> --- a/drivers/ata/libata-transport.c
> >>>> +++ b/drivers/ata/libata-transport.c
> >>>> @@ -294,6 +294,7 @@ int ata_tport_add(struct device *parent,
> >>>>      device_enable_async_suspend(dev);
> >>>>      pm_runtime_set_active(dev);
> >>>>      pm_runtime_enable(dev);
> >>>> +    pm_runtime_forbid(dev);
> >>>>
> >>>>      transport_add_device(dev);
> >>>>      transport_configure_device(dev);
> >>
> >>
> >> There is one more minor issue. I cannot get my disk re-dectected at 3Gbps. Here is when I plugged it in
> >> for the very first time after bootup (plain 3.4-rc3 with the above one-line fix):
> > 
> > I did bisect and found that this is a really old regression introduced
> > in 2.6.37-rc1 with below commit.
> > 
> > commit d9027470b88631d0956ac37cdadfdeb9cdcf2c99
> > Author: Gwendal Grignou <gwendal@google.com>
> > Date:   Tue May 25 12:31:38 2010 -0700
> > 
> >     [libata] Add ATA transport class
> > 
> >     This is a scheleton for libata transport class.
> >     All information is read only, exporting information from libata:
> >     - ata_port class: one per ATA port
> >     - ata_link class: one per ATA port or 15 for SATA Port Multiplier
> >     - ata_device class: up to 2 for PATA link, usually one for SATA.
> > 
> >     Signed-off-by: Gwendal Grignou <gwendal@google.com>
> >     Reviewed-by: Grant Grundler <grundler@google.com>
> >     Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
> > 
> > 
> > Here is the patch to fix it.
> > 
> > Gwendal and Grant,
> > 
> > Would you help to review it?
> > 
> > 
> >>From f696daec7ff63e9b3697e8f7ef8f985152667965 Mon Sep 17 00:00:00 2001
> > From: Lin Ming <ming.m.lin@intel.com>
> > Date: Thu, 19 Apr 2012 15:45:51 +0800
> > Subject: [PATCH] libata: clear error mask of old error history
> > 
> > The old error history was cleared in ata_ering_clear().
> > It only sets ATA_EFLAG_OLD_ER eflags, but the err_mask was not cleared.
> > So ata_ering_map() still iterates the old error history.
> > 
> > This causes problem, for example, wrong probe trials count were returned in
> > ata_eh_schedule_probe(), which in turn causes SATA link speed to be slowed down
> > to 1.5Gbps.
> > 
> > Reported-by: Martin Mokrejs <mmokrejs@fold.natur.cuni.cz>
> > Signed-off-by: Lin Ming <ming.m.lin@intel.com>
> > ---
> >  drivers/ata/libata-eh.c |    3 ++-
> >  1 files changed, 2 insertions(+), 1 deletions(-)
> > 
> > diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c
> > index c61316e..4c6f49b 100644
> > --- a/drivers/ata/libata-eh.c
> > +++ b/drivers/ata/libata-eh.c
> > @@ -419,9 +419,10 @@ int ata_ering_map(struct ata_ering *ering,
> >  	return rc;
> >  }
> >  
> > -int ata_ering_clear_cb(struct ata_ering_entry *ent, void *void_arg)
> > +static int ata_ering_clear_cb(struct ata_ering_entry *ent, void *void_arg)
> >  {
> >  	ent->eflags |= ATA_EFLAG_OLD_ER;
> > +	ent->err_mask = 0;
> >  	return 0;
> >  }
> >  
> 
> Confirming this patch fixed my problem. If you would like to improve something, add
> some 1.5Gbp to 3.0Gbps re-negotiation. As you can see belo, I was desperate and plugged
> in the cable just when kernel lowered the speed to 1.5Gbps. It seems it cannot re-negotiate
> to go back up again. Tested 3.4-rc3 with some patches for pciehp from Yinghai, SATA
> hotplug fix (lin Min), and for some reason have still reverted patch
> 486b10b9f43500741cd63a878d0ef23cd87fc66d (just for completeness).
> 
> 
> I plugged in the cable.
> 
> [   75.818463] ata6: exception Emask 0x10 SAct 0x0 SErr 0x4050000 action 0xe frozen
> [   75.818472] ata6: irq_stat 0x00400040, connection status changed
> [   75.818481] ata6: SError: { PHYRdyChg CommWake DevExch }
> [   75.818500] ata6: hard resetting link
> [   76.557738] ata6: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> [   76.558554] ata6.00: ATA-8: ST3000DM001-9YN166, CC4C, max UDMA/133
> [   76.558564] ata6.00: 5860533168 sectors, multi 0: LBA48 NCQ (depth 31/32), AA
> [   76.559365] ata6.00: configured for UDMA/133
> [   76.577645] ata6: EH complete
> [   76.577802] scsi 5:0:0:0: Direct-Access     ATA      ST3000DM001-9YN1 CC4C PQ: 0 ANSI: 5
> [   76.577926] sd 5:0:0:0: [sdc] 5860533168 512-byte logical blocks: (3.00 TB/2.72 TiB)
> [   76.577929] sd 5:0:0:0: [sdc] 4096-byte physical blocks
> [   76.577949] sd 5:0:0:0: Attached scsi generic sg3 type 0
> [   76.577983] sd 5:0:0:0: [sdc] Write Protect is off
> [   76.577986] sd 5:0:0:0: [sdc] Mode Sense: 00 3a 00 00
> [   76.578010] sd 5:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
> [   76.618312]  sdc: sdc1
> [   76.618641] sd 5:0:0:0: [sdc] Attached SCSI disk
> 
> Unplugged the cable.
> 
> [   80.966608] ata6: exception Emask 0x50 SAct 0x0 SErr 0x4090800 action 0xe frozen
> [   80.966617] ata6: irq_stat 0x00400040, connection status changed
> [   80.966625] ata6: SError: { HostInt PHYRdyChg 10B8B DevExch }
> [   80.966643] ata6: hard resetting link
> [   81.710063] ata6: SATA link down (SStatus 0 SControl 300)
> [   84.575097] ata6: hard resetting link
> [   84.915290] ata6: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> [   84.916778] ata6.00: configured for UDMA/133
> [   84.935231] ata6: EH complete
> [   90.042553] ata6: exception Emask 0x10 SAct 0x0 SErr 0x4090000 action 0xe frozen
> [   90.042562] ata6: irq_stat 0x00400040, connection status changed
> [   90.042570] ata6: SError: { PHYRdyChg 10B8B DevExch }
> [   90.042587] ata6: hard resetting link
> [   90.786545] ata6: SATA link down (SStatus 0 SControl 300)
> [   95.779093] ata6: hard resetting link
> [   96.128574] ata6: SATA link down (SStatus 0 SControl 300)
> [   96.128593] ata6: limiting SATA link speed to 1.5 Gbps
> 
> I plugged in the cable before the "SCSI" was disabled.
> 
> [   98.697321] ata6: hard resetting link
> [   99.044242] ata6: SATA link up 1.5 Gbps (SStatus 113 SControl 310)

SATA link can't go back up to 3.0 Gbps.

I can reproduce this issue too.
Will try to fix it.

I have sent out patch "libata: clear error mask of old error history".
Thanks for your test.

Lin Ming

> [   99.045842] ata6.00: configured for UDMA/133
> [   99.064156] ata6: EH complete
> [  141.857958] EXT4-fs (sdc1): mounted filesystem with ordered data mode. Opts: (null)
>