Re: [PATCH] driver core: Postpone DMA tear-down until after devres release for probe failure

From: Robin Murphy <robin.murphy@arm.com>
To: John Garry <john.garry@huawei.com>, Greg KH <gregkh@linuxfoundation.org>
Cc: rafael@kernel.org, linuxarm@huawei.com,
	chenxiang66@hisilicon.com, geert@linux-m68k.org,
	linux-kernel@vger.kernel.org, iommu@lists.linux-foundation.org,
	linux-arm-kernel@lists.infradead.org, robh@kernel.org,
	hch@lst.de
Subject: Re: [PATCH] driver core: Postpone DMA tear-down until after devres release for probe failure
Date: Thu, 11 Apr 2019 12:01:38 +0100	[thread overview]
Message-ID: <198058c6-63c6-ac3c-62b7-438e5e3b3f0e@arm.com> (raw)
In-Reply-To: <d2504066-3d1a-176f-0f7e-7f86431ef88c@huawei.com>

On 11/04/2019 09:50, John Garry wrote:
> On 04/04/2019 12:17, John Garry wrote:
>> On 03/04/2019 10:20, John Garry wrote:
>>> On 03/04/2019 09:14, Greg KH wrote:
>>>> On Wed, Apr 03, 2019 at 09:02:36AM +0100, John Garry wrote:
>>>>> On 28/03/2019 10:08, John Garry wrote:
>>>>>> In commit 376991db4b64 ("driver core: Postpone DMA tear-down until
>>>>>> after
>>>>>> devres release"), we changed the ordering of tearing down the device
>>>>>> DMA
>>>>>> ops and releasing all the device's resources; this was because the
>>>>>> DMA ops
>>>>>> should be maintained until we release the device's managed DMA
>>>>>> memories.
>>>>>>
>>>>>
>>>>> Hi all,
>>>>>
>>>>> A friendly reminder on this patch... I didn't see any update.
>>>>>
>>>>> I thought that it had some importance.
>>>>>
>>>>> Thanks,
>>>>> John
>>>>>
>>>>>> However, we have seen another crash on an arm64 system when a
>>>>>> device driver probe fails:
>>>>>>
>>>>>>   hisi_sas_v3_hw 0000:74:02.0: Adding to iommu group 2
>>>>>>   scsi host1: hisi_sas_v3_hw
>>>>>>   BUG: Bad page state in process swapper/0  pfn:313f5
>>>>>>   page:ffff7e0000c4fd40 count:1 mapcount:0
>>>>>>   mapping:0000000000000000 index:0x0
>>>>>>   flags: 0xfffe00000001000(reserved)
>>>>>>   raw: 0fffe00000001000 ffff7e0000c4fd48 ffff7e0000c4fd48
>>>>>> 0000000000000000
>>>>>>   raw: 0000000000000000 0000000000000000 00000001ffffffff
>>>>>> 0000000000000000
>>>>>>   page dumped because: PAGE_FLAGS_CHECK_AT_FREE flag(s) set
>>>>>>   bad because of flags: 0x1000(reserved)
>>>>>>   Modules linked in:
>>>>>>   CPU: 49 PID: 1 Comm: swapper/0 Not tainted
>>>>>> 5.1.0-rc1-43081-g22d97fd-dirty #1433
>>>>>>   Hardware name: Huawei D06/D06, BIOS Hisilicon D06 UEFI
>>>>>> RC0 - V1.12.01 01/29/2019
>>>>>>   Call trace:
>>>>>>   dump_backtrace+0x0/0x118
>>>>>>   show_stack+0x14/0x1c
>>>>>>   dump_stack+0xa4/0xc8
>>>>>>   bad_page+0xe4/0x13c
>>>>>>   free_pages_check_bad+0x4c/0xc0
>>>>>>   __free_pages_ok+0x30c/0x340
>>>>>>   __free_pages+0x30/0x44
>>>>>>   __dma_direct_free_pages+0x30/0x38
>>>>>>   dma_direct_free+0x24/0x38
>>>>>>   dma_free_attrs+0x9c/0xd8
>>>>>>   dmam_release+0x20/0x28
>>>>>>   release_nodes+0x17c/0x220
>>>>>>   devres_release_all+0x34/0x54
>>>>>>   really_probe+0xc4/0x2c8
>>>>>>   driver_probe_device+0x58/0xfc
>>>>>>   device_driver_attach+0x68/0x70
>>>>>>   __driver_attach+0x94/0xdc
>>>>>>   bus_for_each_dev+0x5c/0xb4
>>>>>>   driver_attach+0x20/0x28
>>>>>>   bus_add_driver+0x14c/0x200
>>>>>>   driver_register+0x6c/0x124
>>>>>>   __pci_register_driver+0x48/0x50
>>>>>>   sas_v3_pci_driver_init+0x20/0x28
>>>>>>   do_one_initcall+0x40/0x25c
>>>>>>   kernel_init_freeable+0x2b8/0x3c0
>>>>>>   kernel_init+0x10/0x100
>>>>>>   ret_from_fork+0x10/0x18
>>>>>>   Disabling lock debugging due to kernel taint
>>>>>>   BUG: Bad page state in process swapper/0  pfn:313f6
>>>>>>   page:ffff7e0000c4fd80 count:1 mapcount:0
>>>>>> mapping:0000000000000000 index:0x0
>>>>>> [   89.322983] flags: 0xfffe00000001000(reserved)
>>>>>>   raw: 0fffe00000001000 ffff7e0000c4fd88 ffff7e0000c4fd88
>>>>>> 0000000000000000
>>>>>>   raw: 0000000000000000 0000000000000000 00000001ffffffff
>>>>>> 0000000000000000
>>>>>>
>>>>>> The crash occurs for the same reason.
>>>>>>
>>>>>> In this case, on the really_probe() failure path, we are still
>>>>>> clearing
>>>>>> the DMA ops prior to releasing the device's managed memories.
>>>>>>
>>>>>> This patch fixes this issue by reordering the DMA ops teardown and 
>>>>>> the
>>>>>> call to devres_release_all() on the failure path.
>>>>>>
>>>>>> Reported-by: Xiang Chen <chenxiang66@hisilicon.com>
>>>>>> Tested-by: Xiang Chen <chenxiang66@hisilicon.com>
>>>>>> Signed-off-by: John Garry <john.garry@huawei.com>
>>>>
>>>> So does this "fix" 376991db4b64?  If so, should this be added to the
>>>> patch and also backported to the stable trees?
>>>
>>> Hi Greg,
>>>
>>> No, I don't think so. I'd say it supplements it. Here I'm trying to fix
>>> up another path in which we tear down the DMA ops prior to releasing the
>>> device's resources.
>>>
>>> I didn't add a fixes tag as 376991db4b64 didn't have one either. It will
>>> need to be backported to stable, I figure the same as 376991db4b64.
>>
> 
> Hi Greg,
> 
> I still think that we should consider merging this patch.

Bah, sorry, I thought I'd replied to this thread already, but apparently 
not :(

I don't have a particularly strong opinion, but at the moment I am 
leaning slightly towards this being a parallel fix for another part of 
09515ef5ddad, rather than a specific fix of the other fix. Either way, 
you can have a:

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

Thanks,
Robin.