[U-Boot] [PATCH 1/2] mtd: vf610_nfc: remove caching of page in buffer

Bill Pringlemeir bpringlemeir at nbsps.com
Tue Apr 7 18:48:50 CEST 2015


On  7 Apr 2015, stefan at agner.ch wrote:

> On 2015-04-07 16:24, Bill Pringlemeir wrote:

>> The OOB patch also significantly decreases UbiFS mounting time in
>> Linux.  I load Linux itself via tftp/network and not using u-boot
>> with nand.  I guess I should try that.

> Is it UBIFS mounting time or is it bad block scanning? Afaik, it
> should speed up the latter significantly, but I don't see why it
> should speed up the former.

It is both,

* no changes.

[0.840632] nand: device found, Manufacturer ID: 0x2c, Chip ID: 0xca
[0.964838] 0x000001040000-0x000010000000 : "root"
.124s

base ubi0 mount is .833585s

* improved READ_OOB

[0.638869] nand: device found, Manufacturer ID: 0x2c, Chip ID: 0xca
[0.707994] 0x000001040000-0x000010000000 : "root"
.069s

base ubi0 mount 1/10s reduction. .738204s


This is for a 256MB device.  The Ubi mount/scan time is not completely
in-significant.

For instance, here is my last run with improved READ_OOB

   [ 0.942538] ubi0: attaching mtd3
   ...
   [ 1.680742] ubi0: background thread "ubi_bgt0d" started, PID 104

This is my 'base ubi0 mount' numbers.

The time is slightly different than what I recorded previously.  I
booted several times without 'READ_OOB' and the times were consistently
'.83xxS'.  It is possible that the initial BBT scan being quicker
altered something; so it is not UBI use of OOB.  I am not sure.  I just
noted that it was ??significantly?? different.

Regards,
Bill Pringlemeir.


More information about the U-Boot mailing list