[lkml]   [2019]   [Nov]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[PATCH v7 00/24] mm/gup: track dma-pinned pages: FOLL_PIN

OK, here is v7, maybe this is the last one. The corresponding git repo
and branch is: pin_user_pages_tracking_v7

Ira, you reviewed the gup_benchmark patches a bit earlier, but I
removed one or two of those review-by tags, due to invasive changes
I made after your review (in response to further reviews).
So could you please reply to any patches you'd like to have
reviewed-by's restoredto, if any? Mainly I'm thinking of
"mm/gup_benchmark: support pin_user_pages() and related calls". Also
various FOLL_LONGTERM vs pin_longterm*() patches.

The following blurb from the v6 cover letter is still applicable, and
I'll repeat it here so it doesn't get lost in the patch blizzard:

Christoph Hellwig has a preference to do things a little differently,
for the devmap cleanup in patch 5 ("mm: devmap: refactor 1-based
refcounting for ZONE_DEVICE pages"). That came up in a different
review thread, because the patch is out for review in two locations.
Here's that review thread:

...and I'm hoping that we can defer that request, because otherwise
it derails this series, which is starting to otherwise look like
it could be ready for 5.5.

Changes since v6:

* Renamed a couple of routines, to get rid of unnecessary leading

__pin_compound_head() --> grab_compound_head()
__record_subpages() --> record_subpages()

* Fixed the error fallback (put_compound_head()) so as to match the fix
in the previous version: need to put back N * GUP_PIN_COUNTING_BIAS
pages, for FOLL_PIN cases.

* Factored out yet another common chunk of code, into a new grab_page()

* Added a missing compound_head() call to put_compound_head().

* [Re-]added Jens Axboe's reviewed-by tag to the fs/io_uring patch.

* Added more reviewed-by's from Jan Kara.

Changes since v5:

* Fixed the refcounting for huge pages: in most cases, it was
only taking one GUP_PIN_COUNTING_BIAS's worth of refs, when it
should have been taking one GUP_PIN_COUNTING_BIAS for each subpage.

(Much thanks to Jan Kara for spotting that one!)

* Renamed user_page_ref_inc() to try_pin_page(), and added a new
try_pin_compound_head(). This definitely improves readability.

* Factored out some more duplication in the FOLL_PIN and FOLL_GET
cases, in gup.c.

* Fixed up some straggling "get_" --> "pin_" references in the comments.

* Added reviewed-by tags.

Changes since v4:

* Renamed put_user_page*() --> unpin_user_page().

* Removed all pin_longterm_pages*() calls. We will use FOLL_LONGTERM
at the call sites. (FOLL_PIN, however, remains an internal gup flag).

This is very nice: many patches just change three characters now:
get_user_pages --> pin_user_pages. I think we've found the right
balance of wrapper calls and gup flags, for the call sites.

* Updated a lot of documentation and commit logs to match the above
two large changes.

* Changed gup_benchmark tests and run_vmtests, to adapt to one less
use case: there is no pin_longterm_pages() call anymore.

* This includes a new devmap cleanup patch from Dan Williams, along
with a rebased follow-up: patches 4 and 5, already mentioned above.

* Fixed patch 10 ("mm/gup: introduce pin_user_pages*() and FOLL_PIN"),
so as to make pin_user_pages*() calls act as placeholders for the
corresponding get_user_pages*() calls, until a later patch fully
implements the DMA-pinning functionality.

Thanks to Jan Kara for noticing that.

* Fixed the implementation of pin_user_pages_remote().

* Further tweaked patch 2 ("mm/gup: factor out duplicate code from four
routines"), in response to Jan Kara's feedback.

* Dropped a few reviewed-by tags due to changes that invalidated

Changes since v3:

* VFIO fix (patch 8): applied further cleanup: removed a pre-existing,
unnecessary release and reacquire of mmap_sem. Moved the DAX vma
checks from the vfio call site, to gup internals, and added comments
(and commit log) to clarify.

* Due to the above, made a corresponding fix to the
pin_longterm_pages_remote(), which was actually calling the wrong
gup internal function.

* Changed put_user_page() comments, to refer to pin*() APIs, rather than
get_user_pages*() APIs.

* Reverted an accidental whitespace-only change in the IB ODP code.

* Added a few more reviewed-by tags.

Changes since v2:

* Added a patch to convert IB/umem from normal gup, to gup_fast(). This
is also posted separately, in order to hopefully get some runtime

* Changed the page devmap code to be a little clearer,
thanks to Jerome for that.

* Split out the page devmap changes into a separate patch (and moved
Ira's Signed-off-by to that patch).

* Fixed my bug in IB: ODP code does not require pin_user_pages()
semantics. Therefore, revert the put_user_page() calls to put_page(),
and leave the get_user_pages() call as-is.

* As part of the revert, I am proposing here a change directly
from put_user_pages(), to release_pages(). I'd feel better if
someone agrees that this is the best way. It uses the more
efficient release_pages(), instead of put_page() in a loop,
and keep the change to just a few character on one line,
but OTOH it is not a pure revert.

* Loosened the FOLL_LONGTERM restrictions in the
__get_user_pages_locked() implementation, and used that in order
to fix up a VFIO bug. Thanks to Jason for that idea.

* Note the use of release_pages() in IB: is that OK?

* Added a few more WARN's and clarifying comments nearby.

* Many documentation improvements in various comments.

* Moved the new pin_user_pages.rst from Documentation/vm/ to
Documentation/core-api/ .

* Commit descriptions: added clarifying notes to the three patches
(drm/via, fs/io_uring, net/xdp) that already had put_user_page()
calls in place.

* Collected all pending Reviewed-by and Acked-by tags, from v1 and v2
email threads.

* Lot of churn from v2 --> v3, so it's possible that new bugs
sneaked in.

NOT DONE: separate patchset is required:

* __get_user_pages_locked(): stop compensating for
buggy callers who failed to set FOLL_GET. Instead, assert
that FOLL_GET is set (and fail if it's not).

Original cover letter (edited to fix up the patch description numbers)

This applies cleanly to linux-next and mmotm, and also to linux.git if
linux-next's commit 20cac10710c9 ("mm/gup_benchmark: fix MAP_HUGETLB
case") is first applied there.

This provides tracking of dma-pinned pages. This is a prerequisite to
solving the larger problem of proper interactions between file-backed
pages, and [R]DMA activities, as discussed in [1], [2], [3], and in
a remarkable number of email threads since about 2017. :)

A new internal gup flag, FOLL_PIN is introduced, and thoroughly
documented in the last patch's Documentation/vm/pin_user_pages.rst.

I believe that this will provide a good starting point for doing the
layout lease work that Ira Weiny has been working on. That's because
these new wrapper functions provide a clean, constrained, systematically
named set of functionality that, again, is required in order to even
know if a page is "dma-pinned".

In contrast to earlier approaches, the page tracking can be
incrementally applied to the kernel call sites that, until now, have
been simply calling get_user_pages() ("gup"). In other words, opt-in by
changing from this:

get_user_pages() (sets FOLL_GET)

to this:
pin_user_pages() (sets FOLL_PIN)

Because there are interdependencies with FOLL_LONGTERM, a similar
conversion as for FOLL_PIN, was applied. The change was from this:

get_user_pages(FOLL_LONGTERM) (also sets FOLL_GET)

to this:
pin_longterm_pages() (sets FOLL_PIN | FOLL_LONGTERM)

Patch summary:

* Patches 1-9: refactoring and preparatory cleanup, independent fixes

* Patch 10: introduce pin_user_pages(), FOLL_PIN, but no functional
changes yet
* Patches 11-16: Convert existing put_user_page() callers, to use the
new pin*()
* Patch 17: Activate tracking of FOLL_PIN pages.
* Patches 18-20: convert various callers
* Patches: 21-23: gup_benchmark and run_vmtests support
* Patch 24: rename put_user_page*() --> unpin_user_page*()


* I've done some overall kernel testing (LTP, and a few other goodies),
and some directed testing to exercise some of the changes. And as you
can see, gup_benchmark is enhanced to exercise this. Basically, I've been
able to runtime test the core get_user_pages() and pin_user_pages() and
related routines, but not so much on several of the call sites--but those
are generally just a couple of lines changed, each.

Not much of the kernel is actually using this, which on one hand
reduces risk quite a lot. But on the other hand, testing coverage
is low. So I'd love it if, in particular, the Infiniband and PowerPC
folks could do a smoke test of this series for me.

Also, my runtime testing for the call sites so far is very weak:

* io_uring: Some directed tests from liburing exercise this, and they pass.
* process_vm_access.c: A small directed test passes.
* gup_benchmark: the enhanced version hits the new gup.c code, and passes.
* infiniband (still only have crude "IB pingpong" working, on a
good day: it's not exercising my conversions at runtime...)
* VFIO: compiles (I'm vowing to set up a run time test soon, but it's
not ready just yet)
* powerpc: it compiles...
* drm/via: compiles...
* goldfish: compiles...
* net/xdp: compiles...
* media/v4l2: compiles...


* Get the block/bio_vec sites converted to use pin_user_pages().

* Work with Ira and Dave Chinner to weave this together with the
layout lease stuff.


[1] Some slow progress on get_user_pages() (Apr 2, 2019):
[2] DMA and get_user_pages() (LPC: Dec 12, 2018):
[3] The trouble with get_user_pages() (Apr 30, 2018):

Dan Williams (1):
mm: Cleanup __put_devmap_managed_page() vs ->page_free()

John Hubbard (23):
mm/gup: pass flags arg to __gup_device_* functions
mm/gup: factor out duplicate code from four routines
mm/gup: move try_get_compound_head() to top, fix minor issues
mm: devmap: refactor 1-based refcounting for ZONE_DEVICE pages
goldish_pipe: rename local pin_user_pages() routine
IB/umem: use get_user_pages_fast() to pin DMA pages
media/v4l2-core: set pages dirty upon releasing DMA buffers
vfio, mm: fix get_user_pages_remote() and FOLL_LONGTERM
mm/gup: introduce pin_user_pages*() and FOLL_PIN
goldish_pipe: convert to pin_user_pages() and put_user_page()
IB/{core,hw,umem}: set FOLL_PIN via pin_user_pages*(), fix up ODP
mm/process_vm_access: set FOLL_PIN via pin_user_pages_remote()
drm/via: set FOLL_PIN via pin_user_pages_fast()
fs/io_uring: set FOLL_PIN via pin_user_pages()
net/xdp: set FOLL_PIN via pin_user_pages()
mm/gup: track FOLL_PIN pages
media/v4l2-core: pin_user_pages (FOLL_PIN) and put_user_page()
vfio, mm: pin_user_pages (FOLL_PIN) and put_user_page() conversion
powerpc: book3s64: convert to pin_user_pages() and put_user_page()
mm/gup_benchmark: use proper FOLL_WRITE flags instead of hard-coding
mm/gup_benchmark: support pin_user_pages() and related calls
selftests/vm: run_vmtests: invoke gup_benchmark with basic FOLL_PIN
mm, tree-wide: rename put_user_page*() to unpin_user_page*()

Documentation/core-api/index.rst | 1 +
Documentation/core-api/pin_user_pages.rst | 233 +++++++++
arch/powerpc/mm/book3s64/iommu_api.c | 12 +-
drivers/gpu/drm/via/via_dmablit.c | 6 +-
drivers/infiniband/core/umem.c | 19 +-
drivers/infiniband/core/umem_odp.c | 13 +-
drivers/infiniband/hw/hfi1/user_pages.c | 4 +-
drivers/infiniband/hw/mthca/mthca_memfree.c | 8 +-
drivers/infiniband/hw/qib/qib_user_pages.c | 4 +-
drivers/infiniband/hw/qib/qib_user_sdma.c | 8 +-
drivers/infiniband/hw/usnic/usnic_uiom.c | 4 +-
drivers/infiniband/sw/siw/siw_mem.c | 4 +-
drivers/media/v4l2-core/videobuf-dma-sg.c | 8 +-
drivers/nvdimm/pmem.c | 6 -
drivers/platform/goldfish/goldfish_pipe.c | 35 +-
drivers/vfio/vfio_iommu_type1.c | 35 +-
fs/io_uring.c | 6 +-
include/linux/mm.h | 195 ++++++-
include/linux/mmzone.h | 2 +
include/linux/page_ref.h | 10 +
mm/gup.c | 553 +++++++++++++++-----
mm/gup_benchmark.c | 74 ++-
mm/huge_memory.c | 44 +-
mm/hugetlb.c | 36 +-
mm/memremap.c | 76 ++-
mm/process_vm_access.c | 28 +-
mm/vmstat.c | 2 +
net/xdp/xdp_umem.c | 4 +-
tools/testing/selftests/vm/gup_benchmark.c | 21 +-
tools/testing/selftests/vm/run_vmtests | 22 +
30 files changed, 1121 insertions(+), 352 deletions(-)
create mode 100644 Documentation/core-api/pin_user_pages.rst


 \ /
  Last update: 2019-11-21 08:16    [W:0.249 / U:9.452 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site