Re: [PATCH v2 0/9] Add support for shared PTEs across processes

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 7/13/22 08:00, David Hildenbrand wrote:
On 08.07.22 21:36, Khalid Aziz wrote:
On 7/8/22 05:47, David Hildenbrand wrote:
On 02.07.22 06:24, Andrew Morton wrote:
On Wed, 29 Jun 2022 16:53:51 -0600 Khalid Aziz <khalid.aziz@xxxxxxxxxx> wrote:

This patch series implements a mechanism in kernel to allow
userspace processes to opt into sharing PTEs. It adds a new
in-memory filesystem - msharefs.

Dumb question: why do we need a new filesystem for this?  Is it not
feasible to permit PTE sharing for mmaps of tmpfs/xfs/ext4/etc files?


IIRC, the general opinion at LSF/MM was that this approach at hand is
makes people nervous and I at least am not convinced that we really want
to have this upstream.

Hi David,

Hi Khalid,


You are right that sharing page tables across processes feels scary, but at the same time threads already share PTEs and
this just extends that concept to processes.

They share a *mm* including a consistent virtual memory layout (VMA
list). Page table sharing is just a side product of that. You could even
call page tables just an implementation detail to produce that
consistent virtual memory layout -- described for that MM via a
different data structure.

Yes, sharing an mm and vma chain does make it different from implementation point of view.


A number of people have commented on potential usefulness of this concept
and implementation.

... and a lot of people raised concerns. Yes, page table sharing to
reduce memory consumption/tlb misses/... is something reasonable to
have. But that doesn't require mshare, as hugetlb has proven.

The design might be useful for a handful of corner (!) cases, but as the
cover letter only talks about memory consumption of page tables, I'll
not care about those. Once these corner cases are explained and deemed
important, we might want to think of possible alternatives to explore
the solution space.

Memory consumption by page tables is turning out to be significant issue. I mentioned one real-world example from a customer where a 300GB SGA on a 512GB server resulted in OOM when 1500+ processes tried to map parts of the SGA into their address space. Some customers are able to solve this issue by switching to hugetlbfs but that is not feasible for every one.


There were concerns raised about being able to make this safe and reliable.
I had agreed to send a
second version of the patch incorporating feedback from last review and LSF/MM, and that is what v2 patch is about. The

Okay, most of the changes I saw are related to the user interface, not
to any of the actual dirty implementation-detail concerns. And the cover
letter is not really clear what's actually happening under the hood and
what the (IMHO) weird semantics of the design imply (as can be seen from
Andrews reply).

Sure, I will add more details to the cover letter next time. msharefs needs more explanation. I will highlight the creation of a new mm struct for mshare regions that is not owned by any process. There was another under-the-hood change that is listed in changelog but could have been highlighted - "Eliminated the need to point vm_mm in original vmas to the newly synthesized mshare mm". How the fields in new mm struct are used helped make this change and could use more details in cover letter.


suggestion to extend hugetlb PMD sharing was discussed briefly. Conclusion from that discussion and earlier discussion
on mailing list was hugetlb PMD sharing is built with special case code in too many places in the kernel and it is
better to replace it with something more general purpose than build even more on it. Mike can correct me if I got that
wrong.

Yes, I pushed for the removal of that yet-another-hugetlb-special-stuff,
and asked the honest question if we can just remove it and replace it by
something generic in the future. And as I learned, we most probably
cannot rip that out without affecting existing user space. Even
replacing it by mshare() would degrade existing user space.

So the natural thing to reduce page table consumption (again, what this
cover letter talks about) for user space (semi- ?)automatically for
MAP_SHARED files is to factor out what hugetlb has, and teach generic MM
code to cache and reuse page tables (PTE and PMD tables should be
sufficient) where suitable.

For reasonably aligned mappings and mapping sizes, it shouldn't be too
hard (I know, locking ...), to cache and reuse page tables attached to
files -- similar to what hugetlb does, just in a generic way. We might
want a mechanism to enable/disable this for specific processes and/or
VMAs, but these are minor details.

And that could come for free for existing user space, because page
tables, and how they are handled, would just be an implementation detail.


I'd be really interested into what the major roadblocks/downsides
file-based page table sharing has. Because I am not convinced that a
mechanism like mshare() -- that has to be explicitly implemented+used by
user space -- is required for that.


I see two parts to what you are suggesting (please correct me if I get this wrong):

1. Implement a generic page table sharing mechanism
2. Implement a way to use this mechanism from userspace

For 1, your suggestion seems to be extract the page table sharing code from hugetlb and make it generic. My approach is to create a special mm struct to host the shared page tables and create a minimal set of changes to simply get PTEs from this special mm struct whenever a shared VMA is accessed. There may be value to extracting hugetlb page table sharing code and recasting it into this framework of special mm struct. I will look some more into it.

As for 2, is it fair to say you are not fond of explicit opt-in from userspace and would rather have the sharing be file based like hugetlb? That is worth considering but is limiting page table sharing to just file objects reasonable? A goal for mshare mechanism was to allow shared objects to be files, anonymous pages, RDMA buffers, whatever. Idea being if you can map it, you can share it with shared page tables. Maybe that is too ambitious a goal and I am open to course correction.

Thanks,
Khalid



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [NTFS 3]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [NTFS 3]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux