Phil Schwan wrote:
>
> Am I correct in seeing that xfs_bmapi() treats argument 3 (block
> number) as pagesize blocks? (If not, the rest of my questions don't
> really apply...)
I don't think so. xfs_bmapi & in fact most of xfs works
with 512-byte blocks. The exception is fs/page_buf.c,
where the I/O paths work on PAGE_SIZE sized buffers.
>
> Is there a rational way to make it treat these as 512 byte disk
> blocks, or am I going to have to hack around this in the code that
> calls into it? If I have to hack around it, is it possible to have a
> page-aligned, page-sized piece of file data sit on (pagesize/512)
> non-contiguous blocks? Even if that's not possible, how can this
> bmapi code handle this situation when a disk is moved to a machine
> with a larger pagesize, where it may now sit on discontiguous blocks?
>
> The reason I ask is because I have a patch that makes lilo load
> kernels from XFS FSes, but I really doubt it'll work yet on
> non-4k-pagesize machines (I guess none of them use lilo, but I hate to
> leave broken interfaces in there regardless).
Yes, I agree: non-4K sized buffers may not work. This is not
a deliberate choice, more like, its not been tested. Non-PAGE_SIZED
buffers won't work.
--------------------------------------------------------------------------
Rajagopal Ananthanarayanan ("ananth")
Member Technical Staff, SGI.
--------------------------------------------------------------------------
|