[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" ph
From: |
Peter Xu |
Subject: |
Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase |
Date: |
Tue, 12 Sep 2023 14:44:19 -0400 |
On Mon, Sep 11, 2023 at 03:48:38PM -0400, Peter Xu wrote:
> On Sat, Sep 09, 2023 at 03:57:44PM +0100, Joao Martins wrote:
> > > Should I continue to treat them as zero pages written with
> > > save_zero_page_to_file ?
> >
> > MCE had already been forward to the guest, so guest is supposed to not be
> > using
> > the page (nor rely on its contents). Hence destination ought to just see a
> > zero
> > page. So what you said seems like the best course of action.
> >
> > > Or should I consider the case of an ongoing compression
> > > use and create a new code compressing an empty page with
> > > save_compress_page() ?
> > >
> > The compress code looks to be a tentative compression (not guaranteed
> > IIUC), so
> > I am not sure it needs any more logic that just adding at the top of
> > ram_save_target_page_legacy() as Peter suggested?
> >
> > > And what about an RDMA memory region impacted by a memory error ?
> > > This is an important aspect.
> > > Does anyone know how this situation is dealt with ? And how it should be
> > > handled
> > > in Qemu ?
> > >
> >
> > If you refer to guest RDMA MRs that is just guest RAM, not sure we are even
> > aware of those from qemu. But if you refer to the RDMA transport that sits
> > below
> > the Qemu file (or rather acts as an implementation of QemuFile), so
> > handling in
> > ram_save_target_page_legacy() already seems to cover it.
>
> I'm also not familiar enough with RDMA, but it looks tricky indeed. AFAIU
> it's leveraging RDMA_CONTROL_COMPRESS for zero pages for now (with
> RDMACompress.value==0), so it doesn't seem to be using generic migration
> protocols.
>
> If we want to fix all places well, one way to consider is to introduce
> migration_buffer_is_zero(), which can be a wrapper for buffer_is_zero() by
> default, but also returns true for poisoned pages before reading the
> buffer. Then we use it in all three places:
>
> - For compression, in do_compress_ram_page()
> - For RDMA, in qemu_rdma_write_one()
Ah, this may not be enough.. sorry.
It seems this is only one path that RDMA will use to save a target page,
for (!rdma->pin_all || !block->is_ram_block) && !block->remote_keys[chunk].
RDMA seems to also possible to merge buffers if virtually continuous
(qemu_rdma_buffer_mergable()), so IIUC it may not trigger an immediate
access to the guest page until later if it finds continuous pages and skip
even more logic. I suspect that's also problematic for poisoned pages so
we should not allow any merged buffer to contain a poisoned page.
Not sure how complicated will it be to fix rdma specifically, copy again
two rdma developers. One option is we state the issue in rdma and fix
non-rdma first. Looks like rdma needs its own fix anyway.
> - For generic migration, in save_zero_page_to_file() (your current patch)
>
> I suppose then all cases will be fixed. We need to make sure we'll always
> use migration_buffer_is_zero() as the 1st thing to call when QEMU wants to
> migrate a target page. Maybe it'll worth a comment above that function.
>
> Thanks,
>
> --
> Peter Xu
--
Peter Xu
- [PATCH 0/1] Qemu crashes on VM migration after an handled memory error, “William Roche, 2023/09/06
- [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, “William Roche, 2023/09/06
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, Joao Martins, 2023/09/06
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, Peter Xu, 2023/09/06
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, William Roche, 2023/09/06
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, Joao Martins, 2023/09/09
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase, Peter Xu, 2023/09/11
- Re: [PATCH 1/1] migration: skip poisoned memory pages on "ram saving" phase,
Peter Xu <=
- [PATCH v2 0/1] Qemu crashes on VM migration after an handled memory error, “William Roche, 2023/09/14
- [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, “William Roche, 2023/09/14
- Re: [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, Zhijian Li (Fujitsu), 2023/09/14
- Re: [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, William Roche, 2023/09/15
- Re: [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, Zhijian Li (Fujitsu), 2023/09/17
- Re: [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, Zhijian Li (Fujitsu), 2023/09/20
- Re: [PATCH v2 1/1] migration: skip poisoned memory pages on "ram saving" phase, William Roche, 2023/09/20
- [PATCH v3 0/1] Qemu crashes on VM migration after an handled memory error, “William Roche, 2023/09/20
- [PATCH v3 1/1] migration: skip poisoned memory pages on "ram saving" phase, “William Roche, 2023/09/20
- Re: [PATCH v2 0/1] Qemu crashes on VM migration after an handled memory error, Peter Xu, 2023/09/14