Re: [PATCH v5] libata: Add ATA_HORKAGE_NO_NCQ_ON_AMD for Samsung 860 and 870 SSD.

[Hans de Goede <hdegoede@redhat.com>]

Hi All,

On 9/1/21 5:16 PM, Kate Hsuan wrote:
> Many users are reporting that the Samsung 860 and 870 SSD are having
> various issues when combined with AMD SATA controllers and only
> completely disabling NCQ helps to avoid these issues.
> 
> Always disabling NCQ for Samsung 860/870 SSDs regardless of the host
> SATA adapter vendor will cause I/O performance degradation with well
> behaved adapters. To limit the performance impact to AMD adapters,
> introduce the ATA_HORKAGE_NO_NCQ_ON_AMD flag to force disable NCQ
> only for these adapters.
> 
> Also, two libata.force parameters (noncqamd and ncqamd) are introduced
> to disable and enable the NCQ for the system which equiped with AMD
> SATA adapter and Samsung 860 and 870 SSDs. The user can determine NCQ
> function to be enabled or disabled according to the demand.
> 
> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=201693
> Signed-off-by: Kate Hsuan <hpa@redhat.com>

Thanks, v5 looks good to me, but in the mean time I've also gotten the
first replies to my request to collect PCI-ids on systems having this
issue and it looks like checking for PCI_VENDOR_ID_AMD might not be
the right thing to do.

Tor Vic who has a very recent motherboard which does not show this
issue reports the following pci-ids:

06:00.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] FCH
SATA Controller [AHCI mode] [1022:7901] (rev 51)
07:00.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] FCH
SATA Controller [AHCI mode] [1022:7901] (rev 51)

Where as we now have a bunch of lspci -nn outputs from different motherboard
with the issue from: https://bugzilla.kernel.org/show_bug.cgi?id=201693 :

00:11.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] [1002:4391] (rev 40)

00:11.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] [1002:4391] (rev 40)

00:11.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] [1002:4391] (rev 40)

00:11.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD/ATI] 
SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] [1002:4391]

00:11.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] [1002:4391]

Notice these are basically all the same; and it seems these are
all from the AMD/ATI period where the motherboars chipset where still
using PCI_VENDOR_ID_ATI as vendor-id.

So it looks like we actually need to disable NCQ for Samsung 860/870
devices when the SATA controller has a vendor-id of PCI_VENDOR_ID_ATI
rather then AMD.

So lets wait a bit for some more info to become available and then
I think we need a v6 with s/AMD/ATI/ (everywhere, also in the name
of the horkage-flag, the cmdline options, etc.).

Regards,

Hans

> ---
> Changes in v5:
> * The libata.force parameters ncqamd and noncqamd are used to enable and
>   disable the NCQ for the systems equiped with AMD SATA adapter and
>   Samsung 860 and 870 SSDs.
> * The character encoding of the patch comment was fixed.
> 
> Changes in v4:
> * A function ata_dev_check_adapter() is added to check the vendor ID of
>   the adapter.
> * ATA_HORKAGE_NONCQ_ON_AMD was modified to ATA_HORKAGE_NO_NCQ_ON_AMD to
>   align with the naming convention.
> * Commit messages were improved according to reviewer comments.
> 
> Changes in v3:
> * ATA_HORKAGE_NONCQ_ON_ASMEDIA_AMD_MARVELL was modified to
>   ATA_HORKAGE_NONCQ_ON_AMD.
> * Codes were fixed to completely disable NCQ on AMD controller.
> 
> ---
>  drivers/ata/libata-core.c | 34 ++++++++++++++++++++++++++++++++--
>  include/linux/libata.h    |  1 +
>  2 files changed, 33 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
> index 3eda3291952b..e2e900085f99 100644
> --- a/drivers/ata/libata-core.c
> +++ b/drivers/ata/libata-core.c
> @@ -2199,6 +2199,25 @@ static void ata_dev_config_ncq_prio(struct ata_device *dev)
> 
>  }
> 
> +static bool ata_dev_check_adapter(struct ata_device *dev,
> +				  unsigned short vendor_id)
> +{
> +	struct pci_dev *pcidev = NULL;
> +	struct device *parent_dev = NULL;
> +
> +	for (parent_dev = dev->tdev.parent; parent_dev != NULL;
> +	     parent_dev = parent_dev->parent) {
> +		if (dev_is_pci(parent_dev)) {
> +			pcidev = to_pci_dev(parent_dev);
> +			if (pcidev->vendor == vendor_id)
> +				return true;
> +			break;
> +		}
> +	}
> +
> +	return false;
> +}
> +
>  static int ata_dev_config_ncq(struct ata_device *dev,
>  			       char *desc, size_t desc_sz)
>  {
> @@ -2217,6 +2236,13 @@ static int ata_dev_config_ncq(struct ata_device *dev,
>  		snprintf(desc, desc_sz, "NCQ (not used)");
>  		return 0;
>  	}
> +
> +	if (dev->horkage & ATA_HORKAGE_NO_NCQ_ON_AMD &&
> +	    ata_dev_check_adapter(dev, PCI_VENDOR_ID_AMD)) {
> +		snprintf(desc, desc_sz, "NCQ (not used)");
> +		return 0;
> +	}
> +
>  	if (ap->flags & ATA_FLAG_NCQ) {
>  		hdepth = min(ap->scsi_host->can_queue, ATA_MAX_QUEUE);
>  		dev->flags |= ATA_DFLAG_NCQ;
> @@ -3951,9 +3977,11 @@ static const struct ata_blacklist_entry ata_device_blacklist [] = {
>  	{ "Samsung SSD 850*",		NULL,	ATA_HORKAGE_NO_NCQ_TRIM |
>  						ATA_HORKAGE_ZERO_AFTER_TRIM, },
>  	{ "Samsung SSD 860*",		NULL,	ATA_HORKAGE_NO_NCQ_TRIM |
> -						ATA_HORKAGE_ZERO_AFTER_TRIM, },
> +						ATA_HORKAGE_ZERO_AFTER_TRIM |
> +						ATA_HORKAGE_NO_NCQ_ON_AMD, },
>  	{ "Samsung SSD 870*",		NULL,	ATA_HORKAGE_NO_NCQ_TRIM |
> -						ATA_HORKAGE_ZERO_AFTER_TRIM, },
> +						ATA_HORKAGE_ZERO_AFTER_TRIM |
> +						ATA_HORKAGE_NO_NCQ_ON_AMD, },
>  	{ "FCCT*M500*",			NULL,	ATA_HORKAGE_NO_NCQ_TRIM |
>  						ATA_HORKAGE_ZERO_AFTER_TRIM, },
> 
> @@ -6108,6 +6136,8 @@ static int __init ata_parse_force_one(char **cur,
>  		{ "ncq",	.horkage_off	= ATA_HORKAGE_NONCQ },
>  		{ "noncqtrim",	.horkage_on	= ATA_HORKAGE_NO_NCQ_TRIM },
>  		{ "ncqtrim",	.horkage_off	= ATA_HORKAGE_NO_NCQ_TRIM },
> +		{ "noncqamd",	.horkage_on	= ATA_HORKAGE_NO_NCQ_ON_AMD },
> +		{ "ncqamd",	.horkage_off	= ATA_HORKAGE_NO_NCQ_ON_AMD },
>  		{ "dump_id",	.horkage_on	= ATA_HORKAGE_DUMP_ID },
>  		{ "pio0",	.xfer_mask	= 1 << (ATA_SHIFT_PIO + 0) },
>  		{ "pio1",	.xfer_mask	= 1 << (ATA_SHIFT_PIO + 1) },
> diff --git a/include/linux/libata.h b/include/linux/libata.h
> index 3fcd24236793..ef1417152ecd 100644
> --- a/include/linux/libata.h
> +++ b/include/linux/libata.h
> @@ -422,6 +422,7 @@ enum {
>  	ATA_HORKAGE_NOTRIM	= (1 << 24),	/* don't use TRIM */
>  	ATA_HORKAGE_MAX_SEC_1024 = (1 << 25),	/* Limit max sects to 1024 */
>  	ATA_HORKAGE_MAX_TRIM_128M = (1 << 26),	/* Limit max trim size to 128M */
> +	ATA_HORKAGE_NO_NCQ_ON_AMD = (1 << 27),	/* Disable NCQ on AMD chipset */
> 
>  	 /* DMA mask for user DMA control: User visible values; DO NOT
>  	    renumber */
> --
> 2.31.1
>