Re: another pmem variant V2

[Christoph Hellwig]

On Wed, Apr 01, 2015 at 07:33:38PM +0000, Elliott, Robert (Server Storage) wrote:
> I triggered a paging error in the memcpy call for a block read
> from system-udevd (actually in a modified memcpy() for the cache
> attribute experiments).  
> 
> 1. This triggered an illegal schedule() call from an atomic context.
> The call trace is shown below.
> 
> 2. memcpy() doesn't provide exception handling or error reporting.
> Some functions like do so, like __copy_user_nocache in
> arch/x85/lib/copy_user_nocache_64.S.  
> 
> Should pmem only use functions that do so, if available on the
> architecture?

We'll need to define an interface for the function to use if it isn't
plain memcpy, which would include that detail.

But I can't see how that memcpy should ever have to handle a page fault,
I'd be curious how your reproduces this issue.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/