Skip to contentSkip to conversation

mtip32xx: mtip32xx: Disable TRIM support

PATCH In review Asai Thambi S P 3 participants 4 messages Apr 12, 2013 – Apr 15, 2013 Block layer

lore
Share & tools
lore
#0 lore raw

PATCH 2/2 mtip32xx: mtip32xx: Disable TRIM support

Temporarily disabling TRIM support until TRIM related issues are addressed in the firmware.

---
 drivers/block/mtip32xx/mtip32xx.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
drivers/block/mtip32xx/mtip32xx.c +2 -0
index d11ea92..53bad62 100644
@@ -1560,10 +1560,12 @@ static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer)
 	}
 #endif
 
+#ifdef MTIP_TRIM /* Disabling TRIM support temporarily */
 	/* Demux ID.DRAT & ID.RZAT to determine trim support */
 	if (port->identify[69] & (1 << 14) && port->identify[69] & (1 << 5))
 		port->dd->trim_supp = true;
 	else
+#endif
 		port->dd->trim_supp = false;
 
 	/* Set the identify buffer as valid. */
Signature · 1 line
1.7.1
maintainer
#1 lore raw
Quoted 3 lines · Asai Thambi S P

Temporarily disabling TRIM support until TRIM related issues are addressed in the firmware.

How serious is this? We do have released kernels out there with the driver, you might want to consider a stable backport too.

Anyway, applied for 3.10.

Signature · 1 line
Jens Axboe
#2 lore raw
Quoted 8 lines · Asai Thambi S P
Quoted 3 lines

Temporarily disabling TRIM support until TRIM related issues are addressed in the firmware.

How serious is this? We do have released kernels out there with the driver, you might want to consider a stable backport too.

It's a performance issue, not data integrity. Since the ATA trim command is non-queueable, IO traffic must halt while the trim is outstanding. A filesystem that is actively trimming sectors is effectively duty cycling the IO path on this part. We are working to shrink the trim command latency while also investigating whether some sort of offline trim is more appropriate.

Given the incremental endurance gain for trim on 34nm SLC vs the severe throughput degradation under some workloads, it seemed prudent to disable the feature.

If you think that warrants a stable backport, just let me or Asai know.

-Sam

#3 lore raw
In reply to Jens Axboe
Quoted 10 lines · Jens Axboe
Quoted 3 lines · Asai Thambi S P

Temporarily disabling TRIM support until TRIM related issues are addressed in the firmware.

How serious is this? We do have released kernels out there with the driver, you might want to consider a stable backport too.

Anyway, applied for 3.10.

Added TRIM support in 3.9 merge window. If this trivial change could be pushed, it would be good. Otherwise, I can backport to stable 3.9 when released.

--
regards,
asai