Re: better SMP support for kprobes

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -


On Thu, Feb 03, 2005 at 08:04:41PM +0530, Prasanna S Panchamukhi wrote:
> [...]
> copy_from_user succeed only if the page is avaliable in the memory, if
> the page is not avaliable in the memory, it page faults. Kprobes handler
> for handling faults is invoked through page fault notifier [...]

For what it's worth, my other concern about this approach is that it
assumes that a simple catchable page fault is the *only* way that such
an attempt may fail.  What if some other kernel version can generate
an oops instead under particularly odd circumstances?  The function
spec says "user context only; may sleep", but we're proposing to ignore
that.  Instead we hope that only one particular failure mode, the one
that we can catch, is ever possible as "punishment" for this.  


- FChE

Attachment: pgp00000.pgp
Description: PGP signature