Re: [RFC][PATCH] PM: Force GFP_NOIO during suspend/resume (was: Re: [linux-pm] Memory allocations in .suspend became very unreliable)
From: Rafael J. Wysocki
Date: Fri Jan 22 2010 - 15:58:37 EST
On Friday 22 January 2010, KOSAKI Motohiro wrote:
> > > Probably we have multiple option. but I don't think GFP_NOIO is good
> > > option. It assume the system have lots non-dirty cache memory and it isn't
> > > guranteed.
> > Basically nothing is guaranteed in this case. However, does it actually make
> > things _worse_?
> Do you mean we don't need to prevent accidental suspend failure?
> Perhaps, I did misunderstand your intention. If you think your patch solve
> this this issue, I still disagree.
No, I don't.
> but If you think your patch mitigate the pain of this issue, I agree it.
That's what I wanted to say really.
> I don't have any reason to oppose your first patch.
> > What _exactly_ does happen without the $subject patch if the
> > system doesn't have non-dirty cache memory and someone makes a GFP_KERNEL
> > allocation during suspend?
> Page allocator prefer to spent lots time for reclaimable memory searching than
> returning NULL. IOW, it can spent time few second if it doesn't have
> reclaimable memory.
> In typical case, OOM killer forcely make enough free memory if the system
> don't have any memory. But under suspending time, oom killer is disabled.
> So, if the caller (probably drivers) call alloc >1000times, the system
> spent lots seconds.
> In this case, GFP_NOIO doesn't help. slowness behavior is caused by
> freeable memory search, not slow i/o.
> However, if strange i/o device makes any i/o slowness, GFP_NOIO might help.
> In this case, please don't ask me about i/o thing. I don't know ;)
OK, thanks for the explanation.
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/