[PATCH] fsl/sata: create a sysfs entry for rx water mark

Jeff Garzik jgarzik at pobox.com
Thu Apr 4 10:57:25 EST 2013


On 03/04/2013 02:20 AM, Qiang Liu wrote:
> Support config RX WATER MARK via sysfs when running at run-time;
> A wrokaround for fix the exception happened to some WD HDD, found on
> WD3000HLFS-01G6U1, WD3000HLFS-01G6U0, some SSD disks. The read performance
> is also regression (about 30%) when use default value.
>
> According to the latest documents, 0x10 is the default value of RX WATER MARK,
> but exception/performance issue happened to some disks mentioned above.
>
> The exception log as below when testing read performance with IOZone:
> ata1.00: exception Emask 0x0 SAct 0x7 SErr 0x800000 action 0x6 frozen
> ata1: SError: { LinkSeq }
> ata1.00: failed command: READ FPDMA QUEUED
> ata1.00: cmd 60/00:00:ff:2c:14/01:00:02:00:00/40 tag 0 ncq 131072 in
> res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
> ata1.00: status: { DRDY }
> ata1.00: failed command: READ FPDMA QUEUED
> ata1.00: cmd 60/00:08:ff:2d:14/01:00:02:00:00/40 tag 1 ncq 131072 in
> res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
> ata1.00: status: { DRDY }
> ata1.00: failed command: WRITE FPDMA QUEUED
> ata1.00: cmd 61/10:10:af:08:6e/00:00:12:00:00/40 tag 2 ncq 8192 out
> res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
> ata1.00: status: { DRDY }
> ata1: hard resetting link
> ata1: Hardreset failed, not off-lined 0
> ata1: Signature Update detected @ 504 msecs
> ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> ata1.00: configured for UDMA/133
> ata1.00: device reported invalid CHS sector 0
> ata1.00: device reported invalid CHS sector 0
> ata1.00: device reported invalid CHS sector 0
> ata1: EH complete
>
> The exception/performance can be resolved when RX WATER MARK value is 0x16.
>
> Signed-off-by: Qiang Liu <qiang.liu at freescale.com>
> ---
>   drivers/ata/sata_fsl.c |   55 ++++++++++++++++++++++++++++++++++++++++++++++++
>   1 files changed, 55 insertions(+), 0 deletions(-)

applied, though it seems advisable to simply set 0x16 than all this mess





More information about the Linuxppc-dev mailing list