Received-SPF: pass (google.com: domain of linux-kernel+bounces-44609-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <c2b85e83-08ac-42cc-9924-bc6ba115b24d@arm.com>
Date: Tue, 30 Jan 2024 12:34:47 +0000
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] ahci: Extend ASM1061 43-bit DMA address quirk to other
 ASM106x parts
To: Damien Le Moal <dlemoal@kernel.org>,
 Lennert Buytenhek <kernel@wantstofly.org>
Cc: Niklas Cassel <cassel@kernel.org>, linux-ide@vger.kernel.org,
 linux-kernel@vger.kernel.org, John Garry <john.g.garry@oracle.com>,
 Joerg Roedel <jroedel@suse.de>, Szuying Chen <chensiying21@gmail.com>,
 Jesse1_Chang@asmedia.com.tw, Richard_Hsu@asmedia.com.tw,
 Chloe_Chen@asmedia.com.tw
References: <ZbjgTmR5FbAnb-Ua@wantstofly.org>
 <6ab581f1-385c-49af-bff3-aacd1cdbe1d8@kernel.org>
 <ZbjkoKotLvDSeJTA@wantstofly.org>
 <2d111b04-4307-4f47-b3a4-208f571d04a8@kernel.org>
Content-Language: en-GB
From: Robin Murphy <robin.murphy@arm.com>
In-Reply-To: <2d111b04-4307-4f47-b3a4-208f571d04a8@kernel.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

On 30/01/2024 12:26 pm, Damien Le Moal wrote:
> On 1/30/24 20:59, Lennert Buytenhek wrote:
>> On Tue, Jan 30, 2024 at 08:46:23PM +0900, Damien Le Moal wrote:
>>
>>>> ASMedia have confirmed that all ASM106x parts currently listed in
>>>> ahci_pci_tbl[] suffer from the 43-bit DMA address limitation that we ran
>>>> into on the ASM1061, and therefore, we need to apply the quirk added by
>>>> commit 20730e9b2778 to the other supported ASM106x parts as well.
>>>>
>>>> Signed-off-by: Lennert Buytenhek <kernel@wantstofly.org>
>>>
>>> I think this needs a cc: stable tag.
>>
>> The commit that is likely responsible for surfacing this issue is
>> 791c2b17fb40 which went into v6.6 -- so would this then be appropriate,
>> or do you think this should be backported to older versions as well?
> 
> Hmmm... given this is a hardware "bug", it seems safer to backport to all stable
> & lts. From what I understand, the device may be doing bad DMA, regardless of
> what the iommu is doing. Niklas ? you followed this more carefully than I did :)

Yes, that would be the case; in practice though it's likely that people 
just don't tend to use these particular controllers in big systems which 
actually have RAM at >43-bit physical addresses.

Thanks,
Robin.

> 
>>
>> Cc: stable@vger.kernel.org # 6.6.x
>>
>>
>>>> ---
>>>>   drivers/ata/ahci.c | 10 +++++-----
>>>>   1 file changed, 5 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
>>>> index d2460fa985b7..da2e74fce2d9 100644
>>>> --- a/drivers/ata/ahci.c
>>>> +++ b/drivers/ata/ahci.c
>>>> @@ -606,13 +606,13 @@ static const struct pci_device_id ahci_pci_tbl[] = {
>>>>   	{ PCI_VDEVICE(PROMISE, 0x3781), board_ahci },   /* FastTrak TX8660 ahci-mode */
>>>>   
>>>>   	/* ASMedia */
>>>> -	{ PCI_VDEVICE(ASMEDIA, 0x0601), board_ahci },	/* ASM1060 */
>>>> -	{ PCI_VDEVICE(ASMEDIA, 0x0602), board_ahci },	/* ASM1060 */
>>>> +	{ PCI_VDEVICE(ASMEDIA, 0x0601), board_ahci_43bit_dma },	/* ASM1060 */
>>>> +	{ PCI_VDEVICE(ASMEDIA, 0x0602), board_ahci_43bit_dma },	/* ASM1060 */
>>>>   	{ PCI_VDEVICE(ASMEDIA, 0x0611), board_ahci_43bit_dma },	/* ASM1061 */
>>>>   	{ PCI_VDEVICE(ASMEDIA, 0x0612), board_ahci_43bit_dma },	/* ASM1061/1062 */
>>>> -	{ PCI_VDEVICE(ASMEDIA, 0x0621), board_ahci },   /* ASM1061R */
>>>> -	{ PCI_VDEVICE(ASMEDIA, 0x0622), board_ahci },   /* ASM1062R */
>>>> -	{ PCI_VDEVICE(ASMEDIA, 0x0624), board_ahci },   /* ASM1062+JMB575 */
>>>> +	{ PCI_VDEVICE(ASMEDIA, 0x0621), board_ahci_43bit_dma },	/* ASM1061R */
>>>> +	{ PCI_VDEVICE(ASMEDIA, 0x0622), board_ahci_43bit_dma },	/* ASM1062R */
>>>> +	{ PCI_VDEVICE(ASMEDIA, 0x0624), board_ahci_43bit_dma },	/* ASM1062+JMB575 */
>>>>   	{ PCI_VDEVICE(ASMEDIA, 0x1062), board_ahci },	/* ASM1062A */
>>>>   	{ PCI_VDEVICE(ASMEDIA, 0x1064), board_ahci },	/* ASM1064 */
>>>>   	{ PCI_VDEVICE(ASMEDIA, 0x1164), board_ahci },   /* ASM1164 */
>