Re: [PATCH] blktrace: readahead support

Re: [PATCH] blktrace: readahead support

[Milton Miller]

Looking at the commit 

Date: Thu, 6 Jul 2006 08:03:28 +0000 (+0200)
X-Git-Url: http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=40359ccb836866435b03a0cb57345002b587d875

> --- a/block/blktrace.c
> +++ b/block/blktrace.c
..
> @@ -79,6 +79,8 @@ static u32 bio_act[3] __read_mostly = { 
>  	(((rw) & (1 << BIO_RW_BARRIER)) >> (BIO_RW_BARRIER - 0))
>  #define trace_sync_bit(rw)	\
>  	(((rw) & (1 << BIO_RW_SYNC)) >> (BIO_RW_SYNC - 1))
> +#define trace_ahead_bit(rw)	\
> +	(((rw) & (1 << BIO_RW_AHEAD)) << (BIO_RW_AHEAD - 0))
>  
>  /*
>   * The worker for the various blk_add_trace*() types. Fills out a


Than doesn't make sense, we are using the bit position in the BIO_RW_
name space twice instead of factoring it out like the other uses.

Looking at  include/linux/bio.h line 147 we find
#define BIO_RW_AHEAD    1

So this is moving it from bit 1 (value 2) to bit 2 (value 4).

I think the shift should be << (2 - BIO_RW_AHEAD).

milton

Re: [PATCH] blktrace: readahead support

[Nathan Scott]

On Mon, Jul 10, 2006 at 12:46:03AM -0500, Milton Miller wrote:
> > +#define trace_ahead_bit(rw)	\
> > +	(((rw) & (1 << BIO_RW_AHEAD)) << (BIO_RW_AHEAD - 0))
> ...
> I think the shift should be << (2 - BIO_RW_AHEAD).

Quite possibly.  As long as it ends up in slot 4 in that array still..
which it will.

cheers.

-- 
Nathan

Re: [PATCH] blktrace: readahead support

[Jens Axboe]

On Mon, Jul 10 2006, Milton Miller wrote:
> Looking at the commit 
> 
> Date: Thu, 6 Jul 2006 08:03:28 +0000 (+0200)
> X-Git-Url: http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=40359ccb836866435b03a0cb57345002b587d875
> 
> > --- a/block/blktrace.c
> > +++ b/block/blktrace.c
> ..
> > @@ -79,6 +79,8 @@ static u32 bio_act[3] __read_mostly = { 
> >  	(((rw) & (1 << BIO_RW_BARRIER)) >> (BIO_RW_BARRIER - 0))
> >  #define trace_sync_bit(rw)	\
> >  	(((rw) & (1 << BIO_RW_SYNC)) >> (BIO_RW_SYNC - 1))
> > +#define trace_ahead_bit(rw)	\
> > +	(((rw) & (1 << BIO_RW_AHEAD)) << (BIO_RW_AHEAD - 0))
> >  
> >  /*
> >   * The worker for the various blk_add_trace*() types. Fills out a
> 
> 
> Than doesn't make sense, we are using the bit position in the BIO_RW_
> name space twice instead of factoring it out like the other uses.
> 
> Looking at  include/linux/bio.h line 147 we find
> #define BIO_RW_AHEAD    1
> 
> So this is moving it from bit 1 (value 2) to bit 2 (value 4).
> 
> I think the shift should be << (2 - BIO_RW_AHEAD).

You are right, it's a little nonsensical right now. I'll get it fixed
up.

-- 
Jens Axboe