| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Pull iommu updates from Joerg Roedel:
- The most intrusive patch is small and changes the default allocation
policy for DMA addresses.
Before the change the allocator tried its best to find an address in
the first 4GB. But that lead to performance problems when that space
gets exhaused, and since most devices are capable of 64-bit DMA these
days, we changed it to search in the full DMA-mask range from the
beginning.
This change has the potential to uncover bugs elsewhere, in the
kernel or the hardware. There is a Kconfig option and a command line
option to restore the old behavior, but none of them is enabled by
default.
- Add Robin Murphy as reviewer of IOMMU code and maintainer for the
dma-iommu and iova code
- Chaning IOVA magazine size from 1032 to 1024 bytes to save memory
- Some core code cleanups and dead-code removal
- Support for ACPI IORT RMR node
- Support for multiple PCI domains in the AMD-Vi driver
- ARM SMMU changes from Will Deacon:
- Add even more Qualcomm device-tree compatible strings
- Support dumping of IMP DEF Qualcomm registers on TLB sync
timeout
- Fix reference count leak on device tree node in Qualcomm driver
- Intel VT-d driver updates from Lu Baolu:
- Make intel-iommu.h private
- Optimize the use of two locks
- Extend the driver to support large-scale platforms
- Cleanup some dead code
- MediaTek IOMMU refactoring and support for TTBR up to 35bit
- Basic support for Exynos SysMMU v7
- VirtIO IOMMU driver gets a map/unmap_pages() implementation
- Other smaller cleanups and fixes
* tag 'iommu-updates-v5.20-or-v6.0' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu: (116 commits)
iommu/amd: Fix compile warning in init code
iommu/amd: Add support for AVIC when SNP is enabled
iommu/amd: Simplify and Consolidate Virtual APIC (AVIC) Enablement
ACPI/IORT: Fix build error implicit-function-declaration
drivers: iommu: fix clang -wformat warning
iommu/arm-smmu: qcom_iommu: Add of_node_put() when breaking out of loop
iommu/arm-smmu-qcom: Add SM6375 SMMU compatible
dt-bindings: arm-smmu: Add compatible for Qualcomm SM6375
MAINTAINERS: Add Robin Murphy as IOMMU SUBSYTEM reviewer
iommu/amd: Do not support IOMMUv2 APIs when SNP is enabled
iommu/amd: Do not support IOMMU_DOMAIN_IDENTITY after SNP is enabled
iommu/amd: Set translation valid bit only when IO page tables are in use
iommu/amd: Introduce function to check and enable SNP
iommu/amd: Globally detect SNP support
iommu/amd: Process all IVHDs before enabling IOMMU features
iommu/amd: Introduce global variable for storing common EFR and EFR2
iommu/amd: Introduce Support for Extended Feature 2 Register
iommu/amd: Change macro for IOMMU control register bit shift to decimal value
iommu/exynos: Enable default VM instance on SysMMU v7
iommu/exynos: Add SysMMU v7 register set
...
|
|
The IOMMU subsystem has already provided an interface to query whether
the IOMMU hardware is enabled for a specific device. This changes the
check from Intel specific intel_iommu_gfx_mapped (globally exported by
the Intel IOMMU driver) to probing the presence of IOMMU on a specific
device using the generic device_iommu_mapped().
This follows commit cca084692394a ("drm/i915: Use per device iommu check")
which converted drm/i915 driver to use device_iommu_mapped().
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Steve Wahl <steve.wahl@hpe.com>
Link: https://lore.kernel.org/r/20220514014322.2927339-3-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
|
|
Exporting the symbols like intel_gtt_* creates some confusion inside
i915 that has symbols named similarly. In an attempt to isolate
platforms needing intel-gtt.ko, commit 7a5c922377b4 ("drm/i915/gt: Split
intel-gtt functions by arch") moved way too much
inside gt/intel_gt_gmch.c, even the functions that don't callout to this
module. Rename the symbols to make the separation clear.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220617230559.2109427-1-lucas.demarchi@intel.com
|
|
In [1], Christoph Hellwig has proposed to remove the wrappers in
include/linux/pci-dma-compat.h.
Some reasons why this API should be removed have been given by Julia
Lawall in [2].
A coccinelle script has been used to perform the needed transformation.
It can be found in [3].
[1]: https://lore.kernel.org/kernel-janitors/20200421081257.GA131897@infradead.org/
[2]: https://lore.kernel.org/kernel-janitors/alpine.DEB.2.22.394.2007120902170.2424@hadrien/
[3]: https://lore.kernel.org/kernel-janitors/20200716192821.321233-1-christophe.jaillet@wanadoo.fr/
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
|
|
Don't include stuff on behalf of users if they're not strictly necessary
for the header.
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/7bcaa1684587b9b008d3c41468fb40e63c54fbc7.1636977089.git.jani.nikula@intel.com
|
|
In intel_gtt_setup_scratch_page(), pointer "page" is not released if
pci_dma_mapping_error() return an error, leading to a memory leak on
module initialisation failure. Simply fix this issue by freeing "page"
before return.
Fixes: 0e87d2b06cb46 ("intel-gtt: initialize our own scratch page")
Signed-off-by: Qiushi Wu <wu000273@umn.edu>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200522083451.7448-1-chris@chris-wilson.co.uk
|
|
DMA_MASK bit values are different for different generations.
This will become more difficult to manage over time with the open
coded usage of different versions of the device.
Fix by:
disallow setting of dma mask in AGP path (< GEN(5) for i915,
add dma_mask_size to the device info configuration,
updating open code call sequence to the latest interface,
refactoring into a common function for setting the dma segment
and mask info
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
cc: Brian Welty <brian.welty@intel.com>
cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20200417195107.68732-1-michael.j.ruhl@intel.com
|
|
After changing the timing between GTT updates and execution on the GPU,
we started seeing sporadic failures on Ironlake. These were narrowed
down to being an insufficiently strong enough barrier/delay after
updating the GTT and scheduling execution on the GPU. By forcing the
uncached read, and adding the missing barrier for the singular
insert_page (relocation paths), the sporadic failures go away.
Fixes: 983d308cb8f6 ("agp/intel: Serialise after GTT updates")
Fixes: 3497971a71d8 ("agp/intel: Flush chipset writes after updating a single PTE")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Andi Shyti <andi.shyti@intel.com>
Cc: stable@vger.kernel.org # v4.0+
Link: https://patchwork.freedesktop.org/patch/msgid/20200410083535.25464-1-chris@chris-wilson.co.uk
|
|
ioremap has provided non-cached semantics by default since the Linux 2.6
days, so remove the additional ioremap_nocache interface.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
|
|
Keeps things consistent now that we make use of struct resource. This
should keep us covered in case we ever get huge amounts of stolen
memory.
v2: bunch of missing conversions (Chris)
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-10-matthew.auld@intel.com
|
|
Now that we are using struct resource to track the stolen region, it is
more convenient if we track dsm in a resource as well.
v2: check range_overflow when writing to 32b registers (Chris)
pepper in some comments (Chris)
v3: refit i915_stolen_to_dma()
v4: kill ggtt->stolen_size
v5: some more polish
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-6-matthew.auld@intel.com
|
|
Before accessing the GGTT we must flush the PTE writes and make them
visible to the chipset, or else the indirect access may end up in the
wrong page. In commit 3497971a71d8 ("agp/intel: Flush chipset writes
after updating a single PTE"), we noticed corruption of the uploads for
pwrite and for capturing GPU error states, but it was presumed that the
explicit calls to intel_gtt_chipset_flush() were sufficient for the
execbuffer path. However, we have not been flushing the chipset between
the PTE writes and access via the GTT itself.
For simplicity, do the flush after any PTE update rather than try and
batch the flushes on a just-in-time basis.
References: 3497971a71d8 ("agp/intel: Flush chipset writes after updating a single PTE")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171208214616.30147-1-chris@chris-wilson.co.uk
|
|
set_memory_* functions have moved to set_memory.h. Switch to this
explicitly.
Link: http://lkml.kernel.org/r/1488920133-27229-7-git-send-email-labbott@redhat.com
Signed-off-by: Laura Abbott <labbott@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Move the intel_fake_agp_sizes array into the same #ifdef block as it is
used to avoid instantiation when not used, and so triggering a compiler
warning
drivers/char/agp/intel-gtt.c:335:42: warning: ‘intel_fake_agp_sizes’
defined but not used [-Wunused-const-variable=]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170121182233.30852-1-chris@chris-wilson.co.uk
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
|
|
Stolen memory is a hardware resource of known size, so use an accurate
fixed integer type rather than the ambiguous variable size_t. This was
motivated by the next patch spotting inconsistencies in our types.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170106152013.24684-3-chris@chris-wilson.co.uk
|
|
After we update one PTE for a page, the caller expects to be able to
immediately use that through a GGTT read/write. To comply with the
callers expectations we therefore need to flush the chipset buffers
before returning.
Reported-by: Matti Hämäläinen <ccr@tnsp.org>
Fixes: d6473f566417 ("drm/i915: Add support for mapping an object page...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Tested-by: Matti Hämäläinen <ccr@tnsp.org>
Cc: drm-intel-fixes@lists.freedesktop.org
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160818161718.27187-2-chris@chris-wilson.co.uk
|
|
Introduced a new vm specfic callback insert_page() to program a single pte in
ggtt or ppgtt. This allows us to map a single page in to the mappable aperture
space. This can be iterated over to access the whole object by using space as
meagre as page size.
v2: Added low level rpm assertions to insert_page routines (Chris)
v3: Added POSTING_READ post register write (Tvrtko)
v4: Rebase (Ankit)
v5: Removed wmb() and FLUSH_CTL from insert_page, caller to take care
of it (Chris)
v6: insert_page not working correctly without FLSH_CNTL write, added the
write again.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
|
|
The fake agp driver for the intel graphics gart is only needed for ums
support. And we ditched that a long time ago:
commit 03dae59c72ffffd8ef6e005f48ba356c863e0587
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Wed Jul 23 16:27:25 2014 +0200
drm/i915: Ditch UMS config option
With this there's no longer the problem that 2 drivers (fake agp
driver and the drm/i915 driver) fight over the same piece, which fixes
apparent dma leaks detected by CONFIG_DMA_API_DEBUG.
Note that the leak isn't real since intel-gtt refcounts and will tear
down eventually. But the debug code assumes that when the i915 driver
unbinds from the pci device everything should be gone. Which isn't the
case if we have intel-agp enabled - userspace might need it. But by
ditching this intel-gtt setup and teardown is completely tied to the
livetime of the "real" driver.
While at it untangle the init ordering a bit - the fake agp wouldn't
be initialized correctly if i915.ko loads first. Which isn't a problem
since when i915 loads in kms mode you won't need the fake agp support
needed by the ums driver ...
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93793
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1453901881-26425-3-git-send-email-daniel.vetter@ffwll.ch
|
|
Recently discovered by enabling CONFIG_DMA_API_DEBUG in our CI. By the
looks of it broken since forever.
v2: Don't forget to set the scratch page back to wb (Chris). Reuse
intel_gtt_teardown_scratch_page for that (and fix it up to treat
needs_dmar y/n correctly).
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93793
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1453901881-26425-1-git-send-email-daniel.vetter@ffwll.ch
|
|
Backmerge fixes since it's getting out of hand again with the massive
split due to atomic between -next and 4.2-rc. All the bugfixes in
4.2-rc are addressed already (by converting more towards atomic
instead of minimal duct-tape) so just always pick the version in next
for the conflicts in modeset code.
All the other conflicts are just adjacent lines changed.
Conflicts:
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_gem_gtt.c
drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_drv.h
drivers/gpu/drm/i915/intel_ringbuffer.h
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
|
|
In needs_ilk_vtd_wa(), we pass in the GPU device but compared it against
the ids for the mobile GPU and the mobile host bridge. That latter is
impossible and so likely was just a typo for the desktop GPU device id
(which is also buggy).
Fixes commit da88a5f7f7d434e2cde1b3e19d952e6d84533662
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Wed Feb 13 09:31:53 2013 +0000
drm/i915: Disable WC PTE updates to w/a buggy IOMMU on ILK
Reported-by: Ting-Wei Lan <lantw44@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91127
References: https://bugzilla.freedesktop.org/show_bug.cgi?id=60391
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
|
|
We can have exactly 4GB sized ppgtt with 32bit system.
size_t is inadequate for this.
v2: Convert a lot more places (Daniel)
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
An interesting bug occurs on Pineview through which the root cause is
that the writes of the PTE values into the GTT is not serialised with
subsequent memory access through the GTT (when using WC updates of the
PTE values). This is despite there being a posting read after the GTT
update. However, by changing the address of the posting read, the memory
access is indeed serialised correctly.
Whilst we are manipulating the memory barriers, we can remove the
compiler :memory restraint on the intermediate PTE writes knowing that
we explicitly perform a posting read afterwards.
v2: Replace posting reads with explicit write memory barriers - in
particular this is advantages in case of single page objects. Update
comments to mention this issue is only with WC writes.
Testcase: igt/gem_exec_big #pnv
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88191
Tested-by: huax.lu@intel.com (v1)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
- Remove soon-to-be-dead @redhat address.
- Jeff Hartmann wrote the bulk of the original backend code, and should
at least get a mention in the MODULE_AUTHOR for backend.o
- Various people at Intel have done a lot more work than myself on the
intel-* drivers, so again, mention that.
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
If a page isn't allocated as __GFP_MOVEABLE it won't move around, so
no need to grab a reference to lock it into place.
Discovered while reviewing page allocation handling in i915 gem.
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
|
|
None of these files are actually using any __init type directives
and hence don't need to include <linux/init.h>. Most are just a
left over from __devinit and __cpuinit removal, or simply due to
code getting copied from one driver to the next.
Cc: David Airlie <airlied@linux.ie>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Kukjin Kim <kgene.kim@samsung.com>
Cc: Corey Minyard <minyard@acm.org>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Peter Huewe <peterhuewe@gmx.de>
Cc: Ashley Lai <ashley@ashleylai.com>
Cc: Marcel Selhorst <tpmdd@selhorst.net>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Pull drm updates from Dave Airlie:
"Been a bit busy, first week of kids school, and waiting on other trees
to go in before I could send this, so its a bit later than I'd
normally like.
Highlights:
- core:
timestamp fixes, lots of misc cleanups
- new drivers:
bochs virtual vga
- vmwgfx:
major overhaul for their nextgen virt gpu.
- i915:
runtime D3 on HSW, watermark fixes, power well work, fbc fixes,
bdw is no longer prelim.
- nouveau:
gk110/208 acceleration, more pm groundwork, old overlay support
- radeon:
dpm rework and clockgating for CIK, pci config reset, big endian
fixes
- tegra:
panel support and DSI support, build as module, prime.
- armada, omap, gma500, rcar, exynos, mgag200, cirrus, ast:
fixes
- msm:
hdmi support for mdp5"
* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (595 commits)
drm/nouveau: resume display if any later suspend bits fail
drm/nouveau: fix lock unbalance in nouveau_crtc_page_flip
drm/nouveau: implement hooks for needed for drm vblank timestamping support
drm/nouveau/disp: add a method to fetch info needed by drm vblank timestamping
drm/nv50: fill in crtc mode struct members from crtc_mode_fixup
drm/radeon/dce8: workaround for atom BlankCrtc table
drm/radeon/DCE4+: clear bios scratch dpms bit (v2)
drm/radeon: set si_notify_smc_display_change properly
drm/radeon: fix DAC interrupt handling on DCE5+
drm/radeon: clean up active vram sizing
drm/radeon: skip async dma init on r6xx
drm/radeon/runpm: don't runtime suspend non-PX cards
drm/radeon: add ring to fence trace functions
drm/radeon: add missing trace point
drm/radeon: fix VMID use tracking
drm: ast,cirrus,mgag200: use drm_can_sleep
drm/gma500: Lock struct_mutex around cursor updates
drm/i915: Fix the offset issue for the stolen GEM objects
DRM: armada: fix missing DRM_KMS_FB_HELPER select
drm/i915: Decouple GPU error reporting from ring initialisation
...
|
|
In i810_setup(), i830_setup(), and i9xx_setup(), we use the result of
pci_bus_address() as an argument to ioremap() and to compute gtt_phys_addr.
These should use pci_resource_start() instead because we want the CPU
physical address, not the bus address.
If there were an AGP device behind a host bridge that translated addresses,
e.g., a PNP0A08 device with _TRA != 0, this would fix a bug. I'm not aware
of any of those, but they are possible.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Per the Intel 915G/915GV/... Chipset spec (document number 301467-005),
GTTADR is a standard PCI BAR.
The PCI core reads GTTADR at enumeration-time. Use pci_bus_address()
instead of reading it again in the driver. This works correctly for both
32-bit and 64-bit BARs. The spec above only mentions 32-bit GTTADR, but we
should still use the standard interface.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Per the Intel 915G/915GV/... Chipset spec (document number 301467-005),
MMADR is a standard PCI BAR.
The PCI core reads MMADR at enumeration-time. Use pci_bus_address()
instead of reading it again in the driver. This works correctly for both
32-bit and 64-bit BARs. The spec above only mentions 32-bit MMADR, but we
should still use the standard interface.
Also, stop clearing the low 19 bits of the bus address because it's invalid
to use addresses outside the region defined by the BAR. The spec claims
MMADR is 512KB; if that's the case, those bits will be zero anyway.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Per the Intel 915G/915GV/... Chipset spec (document number 301467-005),
GMADR is a standard PCI BAR.
The PCI core reads GMADR at enumeration-time. Use pci_bus_address()
instead of reading it again in the driver. This works correctly for both
32-bit and 64-bit BARs. The spec above only mentions 32-bit GMADR, but
Yinghai's patch (link below) indicates some devices have a 64-bit GMADR.
[bhelgaas: reworked starting from http://lkml.kernel.org/r/1385851238-21085-13-git-send-email-yinghai@kernel.org]
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
The only use of gtt_bus_addr is as an argument to ioremap(), so it is a CPU
physical address, not a bus address. Rename it to gtt_phys_addr to reflect
this.
No functional change.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
We only depend on the intel-gtt module for GTT frobbign on older gens.
The intel_agp module is optional, except for UMS and some old XvMC
userland on gen3. So make AGP support optional. As before, we will
fail the i915 init for UMS and gen3 KMS the same as before if
intel_agp isn't around.
intel-gtt.c is left with a somewhat ugly ifdef mess, but I'm going
to save that for a later cleaning.
At least my gen2 still works with the patch and CONFIG_AGP=n.
v2: Make i915 depend on X86 and PCI, and intel-gtt depend on PCI
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Whilst IOMMU is enabled for the Intel GPU on Ironlake, it appears that
using WC writes to update the PTE on the GPU fails miserably. The
result looks like the majority of the writes do not land leading to
lots of screen corruption and a hard system hang.
v2: s/</<=/ to preserve the current exclusion of Sandybridge
Reported-by: Nathan Myers <ncm@cantrip.org>
Bugzilla: https://bugzilla.freedesktop.org/show_bug.cgi?id=60391
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Nathan Myers <ncm@cantrip.org>
[danvet: Remove cc: stable and add tested-by.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
When I refactored the code initially, I forgot that gen2 uses a
different bar for the CPU mappable aperture. The agp-less code knows
nothing of generations less than 5, so we have to expand the gtt_probe
function to include the mappable base and end.
It was originally broken by me:
commit baa09f5fd8a6d033ec075355dda99a65b7f6a0f3
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Thu Jan 24 13:49:57 2013 -0800
drm/i915: Add probe and remove to the gtt ops
Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
With the probe call in our dispatch table, we can now cut away the
last three remaining members in the intel_gtt shared struct and so
remove it completely.
v2: Rebased on top of Daniel's series
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: bikeshed commit message a bit.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
It is no longer used in the i915 code, so isolate it from the shared
struct.
This was originally part of:
commit 0e275518f325418d559c05327775bff894b237f7
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Mon Jan 14 13:35:33 2013 -0800
agp/intel: decouple more of the agp-i915 sharing
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
That commit had some other hunks which can't be used due to issues
Daniel found in previous commits.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: drop squash notice from the commit since it's imo ok to keep
this one separate.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
The reasoning behind our code taking two paths depending upon whether or
not we may have been configured for IOMMU isn't clear to me. It should
always be safe to use the pci mapping functions as they are designed to
abstract the decision we were handling in i915.
Aside from simpler code, removing another member for the intel_gtt
struct is a nice motivation.
I ran this by Chris, and he wasn't concerned about the extra kzalloc,
and memory references vs. page_to_phys calculation in the case without
IOMMU.
v2: Update commit message
v3: Remove needs_dmar addition from Zhenyu upstream
This reverts (and then other stuff)
commit 20652097dadd9a7fb4d652f25466299974bc78f9
Author: Zhenyu Wang <zhenyuw@linux.intel.com>
Date: Thu Dec 13 23:47:47 2012 +0800
drm/i915: Fix missed needs_dmar setting
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> (v2)
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Squash in follow-up fix to remove the bogus hunk which
deleted the dma_mask configuration for gen6+.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
We already had a mapping in both (minus the phys_addr in AGP).
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
And, move it to where the rest of the logic is.
There is some slight functionality changes. There was extra paranoid
checks in AGP code making sure we never do idle maps on gen2 parts. That
was not duplicated as the simple PCI id check should do the right thing.
v2: use IS_GEN5 && IS_MOBILE check instead. For now, this is the same as
IS_IRONLAKE_M but is more future proof. The workaround docs hint that
more than one platform may be effected, but we've never seen such a
platform in the wild. (Rodrigo, Daniel)
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> (v1)
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
This removes an unused field from the AGP structure and moves it into
the dev_priv structure (with a slightly better name). This builds upon
the kill-agp series already merged.
GSM is a well defined term in the bspec:
GSM: Graphics Stolen Memory
GTT stolen space is defined for storage of the GFX GTT entries in
physical memory. IA can not access GSM directly , it can only access via
GTTMMADR. GT can access GSM directly or through GTTMMADR.
This is not the entire stolen space.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
v2: Accidently removed an ILK case in i9xx_setup (Nicely found by Chris)
CC: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by [v1] : Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
As a quick hack we make the old intel_gtt structure mutable so we can
fool a bunch of the existing code which depends on elements in that data
structure. We can/should try to remove this in a subsequent patch.
This should preserve the old gtt init behavior which upon writing these
patches seems incorrect. The next patch will fix these things.
The one exception is VLV which doesn't have the preserved flush control
write behavior. Since we want to do that for all GEN6+ stuff, we'll
handle that in a later patch. Mainstream VLV support doesn't actually
exist yet anyway.
v2: Update the comment to remove the "voodoo"
Check that the last pte written matches what we readback
v3: actually kill cache_level_to_agp_type since most of the flags will
disappear in an upcoming patch
v4: v3 was actually not what we wanted (Daniel)
Make the ggtt bind assertions better and stricter (Chris)
Fix some uncaught errors at gtt init (Chris)
Some other random stuff that Chris wanted
v5: check for i==0 in gen6_ggtt_bind_object to shut up gcc (Ben)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by [v4]: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Make the cache_level -> agp_flags conversion for pre-gen6 a
tad more robust by mapping everything != CACHE_NONE to the cached agp
flag - we have a 1:1 uncached mapping, but different modes of
cacheable (at least on later generations). Suggested by Chris Wilson.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
It doesn't work since the gtt pte range sits in the middle of the mmio
bar. We didn't notice that since both my and Chris' gen2 machines
don't support PAT and hence all wc io mapping request will
automatically be demoted to uc.
This regression has been introduce in
commit edef7e685da05c13cce50c0126189c80fe2c8f71
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Fri Sep 14 11:57:47 2012 +0100
agp/intel: Use a write-combining map for updating PTEs
Reported-by: Egbert Eich <eich@pdx.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=55834
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Rewriting the PTE entries using an WC mapping is roughly an order of
magnitude faster than through the uncached mapping. This makes an
observable difference on workloads that cycle through large numbers of
buffers, for example Chromium using ShmPixmaps where virtually all the
CPU time is currently spent rebinding the userptr.
v2: Limit the WC mapping to older generations as we have observed that
the TLB invalidation on SandyBridge+ is unreliable with WC updates.
See i-g-t/tests/gem_gtt_cpu_tlb
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Rather than have multiple data structures for describing our page layout
in conjunction with the array of pages, we can migrate all users over to
a scatterlist.
One major advantage, other than unifying the page tracking structures,
this offers is that we replace the vmalloc'ed array (which can be up to
a megabyte in size) with a chain of individual pages which helps reduce
memory pressure.
The disadvantage is that we then do not have a simple array to iterate,
or to access randomly. The common case for this is in the relocation
processing, which will typically fit within a single scatterlist page
and so be almost the same cost as the simple array. For iterating over
the array, the extra function call could be optimised away, but in
reality is an insignificant cost of either binding the pages, or
performing the pwrite/pread.
v2: Fix drm_clflush_sg() to not invoke wbinvd as well! And fix the
trivial compile error from rebasing.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
They've changed it ... for no apparent reason. Meh.
V2: remove unused 'is_hsw' field.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Also properly indent the HB IDs.
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
VLV is a gen7 device, but we don't currently handle that in the switch.
So add it and write the PTEs correctly.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
The PTE format is similar to SNB, but we don't support an MLC and don't
need chipset flushing.
Note: I have my questions whether this is right, given that MLC died
for snb & ivb, that ivb has grown a L3$ cache instead (which vlv seems
to have, too) and that the LLC bit here isn't actually LLC, but just
means 'snoop cpu caches'.
But I plan to burn this all with the heat of a thousands suns in my
gtt rework, so who cares ;-)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Added note.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
Linus Torvalds
Lu Baolu
Lucas De Marchi
Christophe JAILLET
Jani Nikula
Qiushi Wu
Michael J. Ruhl
Chris Wilson
Christoph Hellwig
Matthew Auld
Laura Abbott
Daniel Vetter
Mika Kuoppala
Dave Jones
Paul Gortmaker
Bjorn Helgaas
Yinghai Lu
Ville Syrjälä
Ben Widawsky
Ben Widawsky
Paulo Zanoni
Jesse Barnes