Re: [PATCH] scan.c

[David Woodhouse <dwmw2@infradead.org>]

joakim.tjernlund@lumentis.se said:
>  hmm, I just save the jiffies from the beginning of
> jffs2_scan_medium() and printk the difference just before I leave that
> function.

The real kernel profiling shows a little more detail, which can sometimes 
be useful. Moving the crc32 routine out of line makes it show up nicely too.

>  :-), but seriously once it's on the flash, what's the probability for
> a random flipping bit?  

There are other ways to screw up your flash contents. It doesn't have to be 
a hardware error. I'd be a _lot_ happier if we could continue to expect 
such things and deal with them.

>  At mount time it will save us the trouble to CRC all the data in the
> inode's, won't it? 

Yeah, but we can avoid the need to do that at mount time _anyway_. We can 
probably also ditch most of build.c -- although that doesn't really seem to 
be showing up very high on the profiles anyway.

--
dwmw2