linux-kernel - Re: [PATCH] PCI / PM: Restore the status of PCI devices across hibernation

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAJZ5v0gBeO2559mVNCO-w9_mgzfbzZB6oRY7xc6T=G2DsXV5ag@mail.gmail.com>
Date:   Sat, 17 Jun 2017 00:23:58 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     Bjorn Helgaas <helgaas@...nel.org>
Cc:     Chen Yu <yu.c.chen@...el.com>,
        "Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
        Bjorn Helgaas <bhelgaas@...gle.com>,
        Len Brown <len.brown@...el.com>,
        Dan Williams <dan.j.williams@...el.com>,
        Rui Zhang <rui.zhang@...el.com>,
        Ying Huang <ying.huang@...el.com>,
        Linux PCI <linux-pci@...r.kernel.org>,
        Linux PM <linux-pm@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] PCI / PM: Restore the status of PCI devices across hibernation

On Fri, Jun 16, 2017 at 9:45 PM, Bjorn Helgaas <helgaas@...nel.org> wrote:
> On Thu, May 25, 2017 at 04:49:07PM +0800, Chen Yu wrote:
>> Currently we saw a lot of "No irq handler" errors during hibernation,
>> which caused the system hang finally:
>>
>> [  710.141581] ata4.00: qc timeout (cmd 0xec)
>> [  710.147135] ata4.00: failed to IDENTIFY (I/O error, err_mask=0x4)
>> [  710.154593] ata4.00: revalidation failed (errno=-5)
>> [  710.468124] ata4: SATA link up 6.0 Gbps (SStatus 133 SControl 300)
>> [  710.477746] do_IRQ: 31.151 No irq handler for vector
>>
>> According to above logs, there is an interrupt triggered and it is
>> dispatched to CPU31 with a vector number 151, but there is no handler
>> for it, thus this irq will not get acked and caused irq flood which kill
>> the system. To be more specific, the 31.151 is an interrupt from the ahci
>> host controller.
>>
>> After some investigation, the reason why this issue is triggered is
>> because the thaw_noirq() function does not restore the MSI/MSIX settings
>> across hibernation.
>>
>> The scenario is illustrated below:
>>
>> 1. Before the hibernation starts, the irq 34 is the handler for the ahci device,
>>    which is binded on cpu31.
>> 2. Hibernation starts, the ahci device is put into low power state.
>> 3. All the nonboot CPUs are put offline, so the irq 34 has to be migrated to
>>    the last alive one - CPU0.
>> 4. After the snapshot has been created, all the nonboot CPUs are brought up again,
>>    the CPU affinity for IRQ 34 remains to be 0.
>> 5. ahci device are put into D0.
>> 6. The snapshot is written to the disk.
>>
>> The issue is triggered in step 6, in theory the ahci interrupt should be
>> delivered to CPU0, however the actually result is that this interrupt is
>> delivered to the original CPU31 instead, which cause the "No irq handler" issue.
>>
>> Ying Huang has has provided a clue that, in step 3 it is possible that the writing
>> to the register might not take effect as the PCI devices have been put suspended.
>> Actually it is true:
>> In step 3, the irq 34 affinity is supposed to be modified from 31 to 0,
>> but actually it did not. In __pci_write_msi_msg(), if the device is already
>> in low power state, the low level msi message entry will not be updated
>> but cached. So in theory during the device restore process, the cached msi
>> modification information should be written back to the hardware, and this
>> is what pci_restore_msi_state() do during normal suspend-resume.
>> But this is not the case for hibernation, pci_restore_msi_state() is not
>> invoked currently, to be more specific, pci_restore_state() is not invoked
>> in pci_pm_thaw_noirq(), although pci_save_state() has saved the necessary
>> pci cached information in pci_pm_freeze_noirq().
>>
>> This patch tries to restore the pci status for the device during hibernation,
>> otherwise the status might be lost across hibernation(for example, the MSI/MSIX
>> message settings), which might cause problems during hibernation.
>>
>> Suggested-by: Ying Huang <ying.huang@...el.com>
>> Suggested-by: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
>> Cc: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
>> Cc: Bjorn Helgaas <bhelgaas@...gle.com>
>> Cc: Len Brown <len.brown@...el.com>
>> Cc: Dan Williams <dan.j.williams@...el.com>
>> Cc: Rui Zhang <rui.zhang@...el.com>
>> Cc: Ying Huang <ying.huang@...el.com>
>> Cc: linux-pci@...r.kernel.org
>> Cc: linux-pm@...r.kernel.org
>> Cc: linux-kernel@...r.kernel.org
>> Signed-off-by: Chen Yu <yu.c.chen@...el.com>
>
> Added a stable tag and applied with Rafael's reviewed-by to pci/pm for
> v4.13, thanks!
>
> pci_restore_state() restores a lot of stuff besides MSI/MSI-X: PCIe
> device, link, slot control, ATS, VC, BARs, ACS, IOV.  I guess I'm a
> little surprised that we haven't noticed more issues if all these
> things were broken.

That's because they weren't broken. :-)

None of them is expected to change over the image creation, which is
why pci_pm_thaw_noirq() didn't call pci_restore_state(), but we
overlooked the fact that taking nonboot CPUs offline changed the
configuration of interrupts that needed to be restored afterward.

So this one is really exceptional.

Thanks,
Rafael