Since the page offset can be greater than the size of a page fix
arm_coherent_iommu_map_page so that it maps to nearest page
boundary.
This both prevents unnecessarily mapping memory we don't need to
map and fixes a issue where the unmap wasn't unmapping this extra
memory.
Change-Id: Iaa69aff7505ee75d1f2e69bb0cda814bc6211bd3
Signed-off-by: Liam Mark <lmark@codeaurora.org>
Signed-off-by: Patrick Daly <pdaly@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
size_t type data should not be held under variable of type
int, as this can truncate large values especially on a
64bit system. Fix it.
Change-Id: I5ad1ab321738772a99920e3fa287bda266cb05ed
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Patrick Daly <pdaly@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
DMA buffers may not have cpu virtual addresses when
allocated using DMA_ATTR_NO_KERNEL_MAPPING attribute.
Page address should therefore be looked up while
searching for such dma bufs.
Also fix the printk format specifier to '%pK' to print
kernel pointers.
Change-Id: I5754cde63ace19aed88624a36186d012309d475b
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
Equip arm to support ARCH_SUPPORTS_DEBUG_PAGEALLOC and
enable it by default.
Having this support means all pages freed at buddy level
would be marked as read-only and would help catch any
scribbling over free page area. Allocated pages mapping
attributes is restored to read/write.
ARCH_SUPPORTS_DEBUG_PAGEALLOC needs FORCE_PAGES which
ensures all kernel mapping are populated with page
granularity and it would ensure page level accesses can
be rightly configured.
Change-Id: I9b23d1fb0b9594cc148b082c104da9a773c26057
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
CONFIG_RODATA allows strict kernel mapping permissions to
be followed and accordingly maps regions as read-only,
not-executable etc. correspondingly. CONFIG_RODATA however
assumes all memory regions to be SECTION_SIZE aligned and
section mapped for performance reasons.
With CONFIG_FORCE_PAGES, we force all kernel mappings as
page mapped thus breaking CONFIG_RODATA.
Provide provision to apply permissions at page (pte)
level, if CONFIG_RODATA does not find section mapping.
Change-Id: I8dbf5c3741836bc63a231d8a471cf0306662993b
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
In arm_iommu_map_sg, currently we map each individual link in the given
scatterlist into the SMMU individually such that they may or may not be
virtually contiguous. However, in most of our use cases we
actually want the entire sg list mapped into the SMMU as a single
contiguous range.
This was missing for arm 32bit, add it.
Change-Id: Ibf67f29a60b8d19e526c4719590f2f473ea9dca5
Signed-off-by: Chintan Pandya <cpandya@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
Currently, the decision on which page protection to use
is limited to writecombine and coherent. Expand to include
non consistent memory.
Change-Id: I7585fe3ce804cf321a5585c3d93deb7a7c95045c
Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
During early init, all dma areas are remapped to PAGE_SIZE
granularity. Since full pmd regions are cleared to be
remapped into PAGE_SIZE, ensure that address range is pmd
size aligned while not crossing memory boundaries.
This would ensure that even if address region is not pmd
aligned, its mapping would not be cleared but factored in to
PAGE_SIZE regions.
Change-Id: Iad4ad7fd6169cdc693d532821aba453465addb7c
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
Consider only those memory blocks with lowmem while
remapping into 4KB regions using CONFIG_FORCE_PAGES
feature.
Change-Id: Ie209fce6c310f911d8cf02d977e226660684a6ab
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
CONFIG_FORCE_PAGES introduces a debug option to mark free
pages as read only in order to trigger a fault if any code
attempts to write to a page on the buddy list.
In order to achieve this it splits all section mappings in
to PAGE_SIZE pages during boot. While doing this split, it
wrongly allocates a pte (of page size) for each pmd. Linux
however in armv7 short descriptor format, uses same second
level pte for 2 consecutive pmds offset by the actual second
level table size. Refer comments in kernel file
arch/arm/include/asm/pgtable-2level.h for details.
Fix this by allocating pte for a section mapped pmd while
providing required offset if pte already exists.
Change-Id: Iadbaa5e71e4b220208a7275bf039a2a413349e42
Signed-off-by: Shiraz Hashim <shashim@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
Lowmem is currently mapped with sections (1MB/2MB) whenever
possible due to TLB performance boosts. Sections cannot easily
be changed at runtime however which makes implementing certain
features annoying. Add an option to map lowmem with 4K patches.
This is intended to be used as a debugging feature and should
NOT be used for performance testing.
Change-Id: I9612a99b8e05a022f5ba7e568f21307cf66b5667
Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
Signed-off-by: Susheel Khiani <skhiani@codeaurora.org>
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
io-pgtable-fast was implemented to achieve
better performance for IOMMU map/un-map. Add
DMA API support that goes through io-pgtable-fast
for 32 bit targets.
Change-Id: I3d0560a4331f6b7b87c70d0885df11d12cb1d6ec
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
[qqzhou@codeaurora.org: Reorganize codes to be more clear]
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
For iommus which do not use stage1 translation, stage2 may not be aware
of the device's dma_mask requirements. Choose the dma_ops such that
allocations are made within the dma_mask region and bounce buffers are
used as necessary.
Change-Id: I63ae3ff9e5ab608892d12db7a813264d234699ee
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
When a device probes and arch_setup_dma_ops is called, check if the
device has an associated default iommu_domain. If it does, setup
a virtual address allocator of the type described by the
default domain's attributes.
Mark arm_iommu_create_mapping and arm_iommu_attach_device as
deprecated, since clients will no longer need to use them now that
the iommu framework will associate them with a default domain.
Change-Id: I0721c038c68f0f48b78d50f9a72418dbaed97172
Signed-off-by: Vijayanand Jitta <vjitta@codeaurora.org>
Signed-off-by: Qingqing Zhou <qqzhou@codeaurora.org>
The SD card clock scaling feature has a dependency on the
MMC_SDHCI_MSM which is not defined. Remove this dependency
so that the SD card clock scaling feature can be enabled.
In addition, fix the SD cards 2.0 initialization caused
by the incorrect clock settings.
Change-Id: I3eaec3546f89b86643b1937764288b3f39578c0e
Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>
The number of ACL entries is currently not sanitized, which
can allow for unreasonably large ACL to be allocated for
later processing. Allowing for this to happen doesn't make much
sense, as there aren't any valid usecases that require a large
ACL, so restrict the number of ACL entries to 16. Also, while we're
here, fix up some of the error handling in a few places.
Change-Id: I5d1371ed180608486bcde1e208a36782986313af
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
The condition that determines if we print an error after
failing to register for the mem-buf message queue checks
the return code against EPROBE_DEFER. This is not correct, as
error codes are negative, so fix the check.
Change-Id: I4ed03bde56c6edf66e0ffed060ec23a87a46be54
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
The mem-buf device can handle memory from anywhere in the
system. This includes memory with a physical address above
32 bits, so configure the DMA mask for the mem-buf device
based on the system configuration.
Change-Id: If91bb6d8044e8e6f47994acd7e50633b46596808
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Return proper value when DEBUG_FS is disabled.
Change-Id: I6d46dcac6c6321e1a19468a81906e5f98976d29b
Signed-off-by: Prakruthi Deepak Heragu <pheragu@codeaurora.org>
Add A619 GPU variant support to deal with different
feature set of A619 GPU.
Change-Id: I4df945af138d03868fc0b26d915145e564d056ad
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
If a request is dequeued, the transfer is cancelled. Give back all
the started requests.
In most scenarios, the function driver dequeues all requests of a
transfer when there's a failure. If the function driver follows this,
then it's fine. If not, then we'd be skipping TRBs at different points
within the dequeue and enqueue pointers, making dequeue/enqueue pointers
useless. To enforce and make sure that we're properly skipping TRBs,
cancel all the started requests and give back all the cancelled requests
to the function drivers.
Change-Id: Ia344a480048383f3fd7c16afbc9f7b0062492879
Signed-off-by: Thinh Nguyen <thinhn@synopsys.com>
Patch-mainline: linux-usb @ 05/03/20, 13:24
Signed-off-by: Mayank Rana <mrana@codeaurora.org>
The value printed on cpu idle exit is a return value from the PSCI call.
To avoid confusion, let's change the text to match the return value.
Change-Id: I8938287fb4d2887aa190af835618d1d0e9f18a99
Signed-off-by: Lina Iyer <ilina@codeaurora.org>
