Re: [tip:perf/uprobes] uprobes, mm, x86: Add the ability to install and remove uprobes breakpoints

[Srikar Dronamraju <srikar@linux.vnet.ibm.com>]

> 
> That said, I think that's true of uprobes too. Why the f*ck would
> uprobes do it's "munmap" operation when we walk the page tables? This
> function was called by more than just the actual unmapping, it was
> called by stuff that wants to zap the pages but leave the mapping
> around.
> 

This was pointed out by Oleg earlier and I had moved the code to
unlink_file_vma.

However by the time unlink_file_vma() is called, the pages would
have been unmapped (in unmap_vmas()) and the task->rss_stat counts
accounted (in zap_pte_range()).

If the exiting process has probepoints, uprobe_munmap() checks if the
breakpoint instruction was around before decrementing the probe count.
This check results in a file backed page being re-read by
uprobe_munmap() and also it cannot find the breakpoint (because we read
a file backed page).

i.e 

1. The task->rss_stat counts gets incremented again because we have read
a page.

2. mm->uprobes_state.count which should have decremented, doesnt get
decremented as uprobe_munmap fails to see the breakpoint.

Hence I had to move back the callback to zap pages so that we do the
cleanup before the task->rss_stat counts are accounted.

That said, Oleg has a in-works patch/idea for removing uprobe_munmap and
mm->uprobes_state.count, which when done, will remove the
uprobe_munmap hook. https://lkml.org/lkml/2012/4/16/594

Please do let me know if you have better ideas to handle this.

-- 
Thanks and Regards
Srikar

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>