mbox series

[v5,00/11] IOMMU memory observability

Message ID 20240222173942.1481394-1-pasha.tatashin@soleen.com (mailing list archive)
Headers show
Series IOMMU memory observability | expand

Message

Pasha Tatashin Feb. 22, 2024, 5:39 p.m. UTC
----------------------------------------------------------------------
Changelog
----------------------------------------------------------------------
v5:
- Synced with v6.8-rc5
- Added: Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
- Added: Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
- Addressed review comments from Robin Murphy:
  Updated the header comment in iommu-pages.h
  Removed __iommu_alloc_pages_node(), invoke
  iommu_alloc_pages_node directly.
  Removed unused: __iommu_alloc_page_node()
  Removed __iommu_free_page()
  Renamed: iommu_free_pages_list() -> iommu_put_pages_list()
  Added missing iommu_put_pages_list() to dma-iommu.c in
  iommu/dma: use iommu_put_pages_list() to releae freelist

v4:
- Synced with v6.8-rc3
- Updated commit log for "iommu: account IOMMU allocated memory" as
  suggested by Michal Koutný
- Added more Acked-bys David Rientjes and Thierry Reding
- Added Tested-by Bagas Sanjaya.

v3:
- Sync with v6.7-rc7
- Addressed comments from David Rientjes: s/pages/page/, added
  unlikely() into the branches, expanded comment for
  iommu_free_pages_list().
- Added Acked-bys: David Rientjes

v2:
- Added Reviewed-by Janne Grunau
- Sync with 6.7.0-rc3
- Separated form the series patches:
vhost-vdpa: account iommu allocations
https://lore.kernel.org/all/20231130200447.2319543-1-pasha.tatashin@soleen.com
vfio: account iommu allocations
https://lore.kernel.org/all/20231130200900.2320829-1-pasha.tatashin@soleen.com
as suggested by Jason Gunthorpe
- Fixed SPARC build issue detected by kernel test robot
- Drop the following patches as they do account iommu page tables:
iommu/dma: use page allocation function provided by iommu-pages.h
iommu/fsl: use page allocation function provided by iommu-pages.h
iommu/iommufd: use page allocation function provided by iommu-pages.h
as suggested by Robin Murphy. These patches are not related to IOMMU
page tables. We might need to do a separate work to support DMA
observability.
- Remove support iommu/io-pgtable-arm-v7s as the 2nd level pages are
under a page size, thanks Robin Murphy for pointing this out.

----------------------------------------------------------------------
Description
----------------------------------------------------------------------
IOMMU subsystem may contain state that is in gigabytes. Majority of that
state is iommu page tables. Yet, there is currently, no way to observe
how much memory is actually used by the iommu subsystem.

This patch series solves this problem by adding both observability to
all pages that are allocated by IOMMU, and also accountability, so
admins can limit the amount if via cgroups.

The system-wide observability is using /proc/meminfo:
SecPageTables:    438176 kB

Contains IOMMU and KVM memory.

Per-node observability:
/sys/devices/system/node/nodeN/meminfo
Node N SecPageTables:    422204 kB

Contains IOMMU and KVM memory in the given NUMA node.

Per-node IOMMU only observability:
/sys/devices/system/node/nodeN/vmstat
nr_iommu_pages 105555

Contains number of pages IOMMU allocated in the given node.

Accountability: using sec_pagetables cgroup-v2 memory.stat entry.

With the change, iova_stress[1] stops as limit is reached:

$ ./iova_stress
iova space:     0T      free memory:   497G
iova space:     1T      free memory:   495G
iova space:     2T      free memory:   493G
iova space:     3T      free memory:   491G

stops as limit is reached.

This series encorporates suggestions that came from the discussion
at LPC [2].
----------------------------------------------------------------------
[1] https://github.com/soleen/iova_stress
[2] https://lpc.events/event/17/contributions/1466
----------------------------------------------------------------------
Previous versions
v1: https://lore.kernel.org/all/20231128204938.1453583-1-pasha.tatashin@soleen.com
v2: https://lore.kernel.org/linux-mm/20231130201504.2322355-1-pasha.tatashin@soleen.com
v3: https://lore.kernel.org/all/20231226200205.562565-1-pasha.tatashin@soleen.com
v4: https://lore.kernel.org/all/20240207174102.1486130-1-pasha.tatashin@soleen.com
----------------------------------------------------------------------

Pasha Tatashin (11):
  iommu/vt-d: add wrapper functions for page allocations
  iommu/dma: use iommu_put_pages_list() to releae freelist
  iommu/amd: use page allocation function provided by iommu-pages.h
  iommu/io-pgtable-arm: use page allocation function provided by
    iommu-pages.h
  iommu/io-pgtable-dart: use page allocation function provided by
    iommu-pages.h
  iommu/exynos: use page allocation function provided by iommu-pages.h
  iommu/rockchip: use page allocation function provided by iommu-pages.h
  iommu/sun50i: use page allocation function provided by iommu-pages.h
  iommu/tegra-smmu: use page allocation function provided by
    iommu-pages.h
  iommu: observability of the IOMMU allocations
  iommu: account IOMMU allocated memory

 Documentation/admin-guide/cgroup-v2.rst |   2 +-
 Documentation/filesystems/proc.rst      |   4 +-
 drivers/iommu/amd/amd_iommu.h           |   8 -
 drivers/iommu/amd/init.c                |  91 ++++++------
 drivers/iommu/amd/io_pgtable.c          |  13 +-
 drivers/iommu/amd/io_pgtable_v2.c       |  20 +--
 drivers/iommu/amd/iommu.c               |  13 +-
 drivers/iommu/dma-iommu.c               |   7 +-
 drivers/iommu/exynos-iommu.c            |  14 +-
 drivers/iommu/intel/dmar.c              |  16 +-
 drivers/iommu/intel/iommu.c             |  47 ++----
 drivers/iommu/intel/iommu.h             |   2 -
 drivers/iommu/intel/irq_remapping.c     |  16 +-
 drivers/iommu/intel/pasid.c             |  18 +--
 drivers/iommu/intel/svm.c               |  11 +-
 drivers/iommu/io-pgtable-arm.c          |  15 +-
 drivers/iommu/io-pgtable-dart.c         |  37 ++---
 drivers/iommu/iommu-pages.h             | 186 ++++++++++++++++++++++++
 drivers/iommu/rockchip-iommu.c          |  14 +-
 drivers/iommu/sun50i-iommu.c            |   7 +-
 drivers/iommu/tegra-smmu.c              |  18 ++-
 include/linux/mmzone.h                  |   5 +-
 mm/vmstat.c                             |   3 +
 23 files changed, 361 insertions(+), 206 deletions(-)
 create mode 100644 drivers/iommu/iommu-pages.h

Comments

David Rientjes March 15, 2024, 9:33 p.m. UTC | #1
On Thu, 22 Feb 2024, Pasha Tatashin wrote:

> Pasha Tatashin (11):
>   iommu/vt-d: add wrapper functions for page allocations
>   iommu/dma: use iommu_put_pages_list() to releae freelist
>   iommu/amd: use page allocation function provided by iommu-pages.h
>   iommu/io-pgtable-arm: use page allocation function provided by
>     iommu-pages.h
>   iommu/io-pgtable-dart: use page allocation function provided by
>     iommu-pages.h
>   iommu/exynos: use page allocation function provided by iommu-pages.h
>   iommu/rockchip: use page allocation function provided by iommu-pages.h
>   iommu/sun50i: use page allocation function provided by iommu-pages.h
>   iommu/tegra-smmu: use page allocation function provided by
>     iommu-pages.h
>   iommu: observability of the IOMMU allocations
>   iommu: account IOMMU allocated memory
> 
>  Documentation/admin-guide/cgroup-v2.rst |   2 +-
>  Documentation/filesystems/proc.rst      |   4 +-
>  drivers/iommu/amd/amd_iommu.h           |   8 -
>  drivers/iommu/amd/init.c                |  91 ++++++------
>  drivers/iommu/amd/io_pgtable.c          |  13 +-
>  drivers/iommu/amd/io_pgtable_v2.c       |  20 +--
>  drivers/iommu/amd/iommu.c               |  13 +-
>  drivers/iommu/dma-iommu.c               |   7 +-
>  drivers/iommu/exynos-iommu.c            |  14 +-
>  drivers/iommu/intel/dmar.c              |  16 +-
>  drivers/iommu/intel/iommu.c             |  47 ++----
>  drivers/iommu/intel/iommu.h             |   2 -
>  drivers/iommu/intel/irq_remapping.c     |  16 +-
>  drivers/iommu/intel/pasid.c             |  18 +--
>  drivers/iommu/intel/svm.c               |  11 +-
>  drivers/iommu/io-pgtable-arm.c          |  15 +-
>  drivers/iommu/io-pgtable-dart.c         |  37 ++---
>  drivers/iommu/iommu-pages.h             | 186 ++++++++++++++++++++++++
>  drivers/iommu/rockchip-iommu.c          |  14 +-
>  drivers/iommu/sun50i-iommu.c            |   7 +-
>  drivers/iommu/tegra-smmu.c              |  18 ++-
>  include/linux/mmzone.h                  |   5 +-
>  mm/vmstat.c                             |   3 +
>  23 files changed, 361 insertions(+), 206 deletions(-)
>  create mode 100644 drivers/iommu/iommu-pages.h
> 

Joerg, is this series anticipated to be queued up in the core branch of 
git.kernel.org/pub/scm/linux/kernel/git/joro/iommu.git so it gets into 
linux-next?

This observability seems particularly useful so that we can monitor and 
alert on any unexpected increases (unbounded memory growth from this 
subsystem has in the past caused us issues before the memory is otherwise 
not observable by host software).

Or are we still waiting on code reviews from some folks that we should 
ping?

Thanks!
Joerg Roedel March 15, 2024, 9:53 p.m. UTC | #2
Hi David,

On Fri, Mar 15, 2024 at 02:33:53PM -0700, David Rientjes wrote:
> Joerg, is this series anticipated to be queued up in the core branch of 
> git.kernel.org/pub/scm/linux/kernel/git/joro/iommu.git so it gets into 
> linux-next?
> 
> This observability seems particularly useful so that we can monitor and 
> alert on any unexpected increases (unbounded memory growth from this 
> subsystem has in the past caused us issues before the memory is otherwise 
> not observable by host software).
> 
> Or are we still waiting on code reviews from some folks that we should 
> ping?

A few more reviews would certainly help, but I will also do a review on
my own. If things are looking good I can merge it into the iommu tree
when 6.9-rc3 is released (which is the usual time I start merging new
stuff).

Regards,

	Joerg
Pasha Tatashin April 3, 2024, 1:18 p.m. UTC | #3
On Fri, Mar 15, 2024 at 5:53 PM Joerg Roedel <joro@8bytes.org> wrote:
>
> Hi David,
>
> On Fri, Mar 15, 2024 at 02:33:53PM -0700, David Rientjes wrote:
> > Joerg, is this series anticipated to be queued up in the core branch of
> > git.kernel.org/pub/scm/linux/kernel/git/joro/iommu.git so it gets into
> > linux-next?
> >
> > This observability seems particularly useful so that we can monitor and
> > alert on any unexpected increases (unbounded memory growth from this
> > subsystem has in the past caused us issues before the memory is otherwise
> > not observable by host software).
> >
> > Or are we still waiting on code reviews from some folks that we should
> > ping?
>
> A few more reviews would certainly help, but I will also do a review on
> my own. If things are looking good I can merge it into the iommu tree
> when 6.9-rc3 is released (which is the usual time I start merging new
> stuff).

Hi Joerg,

Would it make sense to stage this series in an unstable branch to get
more test coverage from the 0-day robots?

Thank you,
Pasha
Linu Cherian April 4, 2024, 12:58 a.m. UTC | #4
On 2024-02-22 at 23:09:26, Pasha Tatashin (pasha.tatashin@soleen.com) wrote:
> ----------------------------------------------------------------------
> Changelog
> ----------------------------------------------------------------------
> v5:
> - Synced with v6.8-rc5
> - Added: Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
> - Added: Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
> - Addressed review comments from Robin Murphy:
>   Updated the header comment in iommu-pages.h
>   Removed __iommu_alloc_pages_node(), invoke
>   iommu_alloc_pages_node directly.
>   Removed unused: __iommu_alloc_page_node()
>   Removed __iommu_free_page()
>   Renamed: iommu_free_pages_list() -> iommu_put_pages_list()
>   Added missing iommu_put_pages_list() to dma-iommu.c in
>   iommu/dma: use iommu_put_pages_list() to releae freelist
> 
> v4:
> - Synced with v6.8-rc3
> - Updated commit log for "iommu: account IOMMU allocated memory" as
>   suggested by Michal Koutný
> - Added more Acked-bys David Rientjes and Thierry Reding
> - Added Tested-by Bagas Sanjaya.
> 
> v3:
> - Sync with v6.7-rc7
> - Addressed comments from David Rientjes: s/pages/page/, added
>   unlikely() into the branches, expanded comment for
>   iommu_free_pages_list().
> - Added Acked-bys: David Rientjes
> 
> v2:
> - Added Reviewed-by Janne Grunau
> - Sync with 6.7.0-rc3
> - Separated form the series patches:
> vhost-vdpa: account iommu allocations
> https://lore.kernel.org/all/20231130200447.2319543-1-pasha.tatashin@soleen.com
> vfio: account iommu allocations
> https://lore.kernel.org/all/20231130200900.2320829-1-pasha.tatashin@soleen.com
> as suggested by Jason Gunthorpe
> - Fixed SPARC build issue detected by kernel test robot
> - Drop the following patches as they do account iommu page tables:
> iommu/dma: use page allocation function provided by iommu-pages.h
> iommu/fsl: use page allocation function provided by iommu-pages.h
> iommu/iommufd: use page allocation function provided by iommu-pages.h
> as suggested by Robin Murphy. These patches are not related to IOMMU
> page tables. We might need to do a separate work to support DMA
> observability.
> - Remove support iommu/io-pgtable-arm-v7s as the 2nd level pages are
> under a page size, thanks Robin Murphy for pointing this out.
> 
> ----------------------------------------------------------------------
> Description
> ----------------------------------------------------------------------
> IOMMU subsystem may contain state that is in gigabytes. Majority of that
> state is iommu page tables. Yet, there is currently, no way to observe
> how much memory is actually used by the iommu subsystem.
> 
> This patch series solves this problem by adding both observability to
> all pages that are allocated by IOMMU, and also accountability, so
> admins can limit the amount if via cgroups.
> 
> The system-wide observability is using /proc/meminfo:
> SecPageTables:    438176 kB
> 
> Contains IOMMU and KVM memory.

Can you please clarify what does KVM memory refers to here ?
Does it mean the VFIO map / virtio-iommu invoked ones for a guest VM?  

> 
> Per-node observability:
> /sys/devices/system/node/nodeN/meminfo
> Node N SecPageTables:    422204 kB
> 
> Contains IOMMU and KVM memory in the given NUMA node.
> 
> Per-node IOMMU only observability:
> /sys/devices/system/node/nodeN/vmstat
> nr_iommu_pages 105555
> 
> Contains number of pages IOMMU allocated in the given node.
> 
> Accountability: using sec_pagetables cgroup-v2 memory.stat entry.
> 
> With the change, iova_stress[1] stops as limit is reached:
> 
> $ ./iova_stress
> iova space:     0T      free memory:   497G
> iova space:     1T      free memory:   495G
> iova space:     2T      free memory:   493G
> iova space:     3T      free memory:   491G
> 
> stops as limit is reached.
> 
> This series encorporates suggestions that came from the discussion
> at LPC [2].
> ----------------------------------------------------------------------
> [1] https://github.com/soleen/iova_stress
> [2] https://lpc.events/event/17/contributions/1466
> ----------------------------------------------------------------------
> Previous versions
> v1: https://lore.kernel.org/all/20231128204938.1453583-1-pasha.tatashin@soleen.com
> v2: https://lore.kernel.org/linux-mm/20231130201504.2322355-1-pasha.tatashin@soleen.com
> v3: https://lore.kernel.org/all/20231226200205.562565-1-pasha.tatashin@soleen.com
> v4: https://lore.kernel.org/all/20240207174102.1486130-1-pasha.tatashin@soleen.com
> ----------------------------------------------------------------------
> 
> Pasha Tatashin (11):
>   iommu/vt-d: add wrapper functions for page allocations
>   iommu/dma: use iommu_put_pages_list() to releae freelist
>   iommu/amd: use page allocation function provided by iommu-pages.h
>   iommu/io-pgtable-arm: use page allocation function provided by
>     iommu-pages.h
>   iommu/io-pgtable-dart: use page allocation function provided by
>     iommu-pages.h
>   iommu/exynos: use page allocation function provided by iommu-pages.h
>   iommu/rockchip: use page allocation function provided by iommu-pages.h
>   iommu/sun50i: use page allocation function provided by iommu-pages.h
>   iommu/tegra-smmu: use page allocation function provided by
>     iommu-pages.h
>   iommu: observability of the IOMMU allocations
>   iommu: account IOMMU allocated memory
> 
>  Documentation/admin-guide/cgroup-v2.rst |   2 +-
>  Documentation/filesystems/proc.rst      |   4 +-
>  drivers/iommu/amd/amd_iommu.h           |   8 -
>  drivers/iommu/amd/init.c                |  91 ++++++------
>  drivers/iommu/amd/io_pgtable.c          |  13 +-
>  drivers/iommu/amd/io_pgtable_v2.c       |  20 +--
>  drivers/iommu/amd/iommu.c               |  13 +-
>  drivers/iommu/dma-iommu.c               |   7 +-
>  drivers/iommu/exynos-iommu.c            |  14 +-
>  drivers/iommu/intel/dmar.c              |  16 +-
>  drivers/iommu/intel/iommu.c             |  47 ++----
>  drivers/iommu/intel/iommu.h             |   2 -
>  drivers/iommu/intel/irq_remapping.c     |  16 +-
>  drivers/iommu/intel/pasid.c             |  18 +--
>  drivers/iommu/intel/svm.c               |  11 +-
>  drivers/iommu/io-pgtable-arm.c          |  15 +-
>  drivers/iommu/io-pgtable-dart.c         |  37 ++---
>  drivers/iommu/iommu-pages.h             | 186 ++++++++++++++++++++++++
>  drivers/iommu/rockchip-iommu.c          |  14 +-
>  drivers/iommu/sun50i-iommu.c            |   7 +-
>  drivers/iommu/tegra-smmu.c              |  18 ++-
>  include/linux/mmzone.h                  |   5 +-
>  mm/vmstat.c                             |   3 +
>  23 files changed, 361 insertions(+), 206 deletions(-)
>  create mode 100644 drivers/iommu/iommu-pages.h
> 
> -- 
> 2.44.0.rc0.258.g7320e95886-goog
>
Pasha Tatashin April 4, 2024, 1:50 p.m. UTC | #5
> > Description
> > ----------------------------------------------------------------------
> > IOMMU subsystem may contain state that is in gigabytes. Majority of that
> > state is iommu page tables. Yet, there is currently, no way to observe
> > how much memory is actually used by the iommu subsystem.
> >
> > This patch series solves this problem by adding both observability to
> > all pages that are allocated by IOMMU, and also accountability, so
> > admins can limit the amount if via cgroups.
> >
> > The system-wide observability is using /proc/meminfo:
> > SecPageTables:    438176 kB
> >
> > Contains IOMMU and KVM memory.
>
> Can you please clarify what does KVM memory refers to here ?
> Does it mean the VFIO map / virtio-iommu invoked ones for a guest VM?

This means that nested page tables that are managed by KVM, and device
page tables that are managed by IOMMU are all accounted in
SecPageTables (secondary page tables). The decision to account them
both in one field of meminfo was made at LPC'23.

Pasha
Joerg Roedel April 12, 2024, 10:10 a.m. UTC | #6
Pasha,

On Thu, Feb 22, 2024 at 05:39:26PM +0000, Pasha Tatashin wrote:
> Pasha Tatashin (11):
>   iommu/vt-d: add wrapper functions for page allocations
>   iommu/dma: use iommu_put_pages_list() to releae freelist
>   iommu/amd: use page allocation function provided by iommu-pages.h
>   iommu/io-pgtable-arm: use page allocation function provided by
>     iommu-pages.h
>   iommu/io-pgtable-dart: use page allocation function provided by
>     iommu-pages.h
>   iommu/exynos: use page allocation function provided by iommu-pages.h
>   iommu/rockchip: use page allocation function provided by iommu-pages.h
>   iommu/sun50i: use page allocation function provided by iommu-pages.h
>   iommu/tegra-smmu: use page allocation function provided by
>     iommu-pages.h
>   iommu: observability of the IOMMU allocations
>   iommu: account IOMMU allocated memory

Some problems with this:

  1. I get DKIM failures when downloading this patch-set with b4, can
     you please send them via a mailserver with working DKIM?

  2. They don't apply to v6.9-rc3. Please rebase to that version and
     are-send.

Thanks,

	Joerg
Pasha Tatashin April 12, 2024, 5:06 p.m. UTC | #7
> Some problems with this:
>
>   1. I get DKIM failures when downloading this patch-set with b4, can
>      you please send them via a mailserver with working DKIM?

I was in the process of migrating from google domains to a different
registrar, but I think now the issue is resolved. I will verify it.

>   2. They don't apply to v6.9-rc3. Please rebase to that version and
>      are-send.

I will.

Thanks,
Pasha