Ensure GPU can make forward progress on an atomic SVM GPU fault by
giving the GPU a timeslice of 5ms
v2:
- Reduce timeslice to 5ms
- Double timeslice on retry
- Split out GPU SVM changes into independent patch
v5:
- Double timeslice in a few more places
Fixes: 2f118c9491 ("drm/xe: Add SVM VRAM migration")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-5-matthew.brost@intel.com
(cherry picked from commit a5d8d3be1d)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Add timeslicing support to GPU SVM which will guarantee the GPU a
minimum execution time on piece of physical memory before migration back
to CPU. Intended to implement strict migration policies which require
memory to be in a certain placement for correct execution.
Required for shared CPU and GPU atomics on certain devices.
Fixes: 99624bdff8 ("drm/gpusvm: Add support for GPU Shared Virtual Memory")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-4-matthew.brost@intel.com
(cherry picked from commit 8dc1812b5b)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Mixing GPU and CPU atomics does not work unless a strict migration
policy of GPU atomics must be device memory. Enforce a policy of must be
in VRAM with a retry loop of 3 attempts, if retry loop fails abort
fault.
Removing always_migrate_to_vram modparam as we now have real migration
policy.
v2:
- Only retry migration on atomics
- Drop alway migrate modparam
v3:
- Only set vram_only on DGFX (Himal)
- Bail on get_pages failure if vram_only and retry count exceeded (Himal)
- s/vram_only/devmem_only
- Update xe_svm_range_is_valid to accept devmem_only argument
v4:
- Fix logic bug get_pages failure
v5:
- Fix commit message (Himal)
- Mention removing always_migrate_to_vram in commit message (Lucas)
- Fix xe_svm_range_is_valid to check for devmem pages
- Bail on devmem_only && !migrate_devmem (Thomas)
v6:
- Add READ_ONCE barriers for opportunistic checks (Thomas)
- Pair READ_ONCE with WRITE_ONCE (Thomas)
v7:
- Adjust comments (Thomas)
Fixes: 2f118c9491 ("drm/xe: Add SVM VRAM migration")
Cc: stable@vger.kernel.org
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-3-matthew.brost@intel.com
(cherry picked from commit a9ac0fa455)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
This commit adds a new flag, devmem_only, to the drm_gpusvm structure. The
purpose of this flag is to ensure that the get_pages function allocates
memory exclusively from the device's memory. If the allocation from
device memory fails, the function will return an -EFAULT error.
Required for shared CPU and GPU atomics on certain devices.
v3:
- s/vram_only/devmem_only/
Fixes: 99624bdff8 ("drm/gpusvm: Add support for GPU Shared Virtual Memory")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-2-matthew.brost@intel.com
(cherry picked from commit 8a9b978ebd)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Add Wa_22021007897 for the Xe2_HPG (graphics version: 20.01) IP. It is
a permanent workaround, and applicable on all the steppings.
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Reviewed-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Signed-off-by: Aradhya Bhatia <aradhya.bhatia@intel.com>
Link: https://lore.kernel.org/r/20250512065004.2576-1-aradhya.bhatia@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit e5c13e2c50)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Add a new compatible string for ax45mp-cache on QiLai SoC.
Also, add allOf constraints to enforce specific cache-sets and cache-size
values for qilai-ax45mp-cache.
Signed-off-by: Ben Zong-You Xie <ben717@andestech.com>
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
On VCN v4.0.5 there is a race condition where the WPTR is not
updated after starting from idle when doorbell is used. Adding
register read-back after written at function end is to ensure
all register writes are done before they can be used.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Tested-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 07c9db090b)
Cc: stable@vger.kernel.org
This reverts commit 272e6aab14.
Applying degamma curve to the cursor by default breaks Linux userspace
expectation.
On Linux, AMD display manager enables cursor degamma ROM just for
implict sRGB on HW versions where degamma is split into two blocks:
degamma ROM for pre-defined TFs and `gamma correction` for user/custom
curves, and degamma ROM settings doesn't apply to cursor plane.
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1513
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2803
Reported-by: Michel Dänzer <michel.daenzer@mailbox.org>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4144
Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit f6a305d474)
Cc: stable@vger.kernel.org
When operating at low speeds, the display may throw an underflow
error and the display itself goes blank. Increasing the fifo-th
value appears to correct this problem and the display can now
operate when the system is operating at speeds as low as 100MHz.
Signed-off-by: Adam Ford <aford173@gmail.com>
Link: https://lore.kernel.org/r/20250504102959.81830-1-aford173@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Expose the debugfs file node for user space to dump the IFWI image
on spirom.
For one transaction between PSP and host, it will read out the
images on both active and inactive partitions so a buffer with two
times the size of maximum IFWI image (currently 16MByte) is needed.
v2: move the vbios gfl macros to the common header and rename the
bo triplet struct to spirom_bo for this specific usage (Hawking)
v3: return directly the result of last command execution (Lijo)
Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As the userq resource was already freed at the drm_release
early phase, it should avoid freeing userq resource again
at the later kms postclose callback.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
A circular locking dependency was detected between the global
`adev->userq_mutex` and per-file `userq_mgr->userq_mutex` when
creating user queues. The issue occurs because:
1. `amdgpu_userq_suspend()` and `amdgpu_userq_resume` take `adev->userq_mutex` first, then
`userq_mgr->userq_mutex`
2. While `amdgpu_userq_create()` takes them in reverse order
This patch resolves the issue by:
1. Moving the `adev->userq_mutex` lock earlier in `amdgpu_userq_create()`
to cover the `amdgpu_userq_ensure_ev_fence()` call
2. Releasing it after we're done with both queue creation and the
scheduling halt check
v2: remove unused adev->userq_mutex lock (Prike)
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On VCN v4.0.5 there is a race condition where the WPTR is not
updated after starting from idle when doorbell is used. Adding
register read-back after written at function end is to ensure
all register writes are done before they can be used.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Tested-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If navi_amd_register_client() fails, the previous i2c_dw_probe() call
should be undone by a corresponding i2c_del_adapter() call, as already done
in the remove function.
Fixes: 17631e8ca2 ("i2c: designware: Add driver support for AMD NAVI GPU")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc: <stable@vger.kernel.org> # v5.13+
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
Link: https://lore.kernel.org/r/fcd9651835a32979df8802b2db9504c523a8ebbb.1747158983.git.christophe.jaillet@wanadoo.fr
This reverts commit 272e6aab14.
Applying degamma curve to the cursor by default breaks Linux userspace
expectation.
On Linux, AMD display manager enables cursor degamma ROM just for
implict sRGB on HW versions where degamma is split into two blocks:
degamma ROM for pre-defined TFs and `gamma correction` for user/custom
curves, and degamma ROM settings doesn't apply to cursor plane.
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1513
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2803
Reported-by: Michel Dänzer <michel.daenzer@mailbox.org>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4144
Signed-off-by: Melissa Wen <mwen@igalia.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Record trace_clock information in the trace_scratch area and recover
the trace_clock when boot, so that reader can docode the timestamp
correctly.
Note that since most trace_clocks records the timestamp in nano-
seconds, this is not a bug. But some trace_clock, like counter and
tsc will record the counter value. Only for those trace_clock user
needs this information.
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/174720625803.1925039.1815089037443798944.stgit@mhiramat.tok.corp.google.com
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The sched_switch and sched_waking events hardcoded the length of the comm
it recorded because these events were created before the dynamic strings
were implemented. Unfortunately, several other events copied this method.
As the size of the comm may change in the future, make the string dynamic.
The dynamic string requires a 4 byte meta data to hold the size and offset
of the string. The amount stored in the ring buffer will then be the
strlen(comm) + 5 (for the \n), and aligned to 4 bytes if there's no other
strings. This means that a task comm can have up to 10 characters before it
requires another 4 bytes in the ring buffer. Most tasks are usually less
than that, so this should not be a problem, and it also allows the name to
be extended over the TASK_COMM_LEN [1]
Note, sched_switch and the sched_waking trace events still hardcode the
length, as there is tooling that still requires that. An effort to update
the tooling will be made to allow this to change in the future.
[1] https://lore.kernel.org/all/20250507110444.963779-1-bhupesh@igalia.com/
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Bhupesh <bhupesh@igalia.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Juri Lelli <juri.lelli@redhat.com>
Link: https://lore.kernel.org/20250507133458.51bafd95@gandalf.local.home
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Most tracepoints in the kernel are created with TRACE_EVENT(). The
TRACE_EVENT() macro (and DECLARE_EVENT_CLASS() and DEFINE_EVENT() where in
reality, TRACE_EVENT() is just a helper macro that calls those other two
macros), will create not only a tracepoint (the function trace_<event>()
used in the kernel), it also exposes the tracepoint to user space along
with defining what fields will be saved by that tracepoint.
There are a few places that tracepoints are created in the kernel that are
not exposed to userspace via tracefs. They can only be accessed from code
within the kernel. These tracepoints are created with DEFINE_TRACE()
Most of these tracepoints end with "_tp". This is useful as when the
developer sees that, they know that the tracepoint is for in-kernel only
(meaning it can only be accessed inside the kernel, either directly by the
kernel or indirectly via modules and BPF programs) and is not exposed to
user space.
Instead of making this only a process to add "_tp", enforce it by making
the DECLARE_TRACE() append the "_tp" suffix to the tracepoint. This
requires adding DECLARE_TRACE_EVENT() macros for the TRACE_EVENT() macro
to use that keeps the original name.
Link: https://lore.kernel.org/all/20250418083351.20a60e64@gandalf.local.home/
Cc: netdev <netdev@vger.kernel.org>
Cc: Jiri Olsa <olsajiri@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: David Ahern <dsahern@kernel.org>
Cc: Juri Lelli <juri.lelli@gmail.com>
Cc: Breno Leitao <leitao@debian.org>
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Link: https://lore.kernel.org/20250510163730.092fad5b@gandalf.local.home
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Instead of find_first_bit() use the dedicated bitmap_empty(),
and make upper_empty() a nice one-liner.
While there, fix opencoded BITS_PER_TYPE().
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/20250429195119.620204-1-yury.norov@gmail.com
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
This commit adds the NVME_QUIRK_NO_DEEPEST_PS quirk for device
[126f:2262], which belongs to device SOLIDIGM P44 Pro SSDPFKKW020X7
The device frequently have trouble exiting the deepest power state (5),
resulting in the entire disk being unresponsive.
Verified by setting nvme_core.default_ps_max_latency_us=10000 and
observing the expected behavior.
Signed-off-by: Ilya Guterman <amfernusus@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
In preparation of hierarchical scheduling support, add @sch to scx_exit()
and friends:
- scx_exit/error() updated to take explicit @sch instead of assuming
scx_root.
- scx_kf_exit/error() added. These are to be used from kfuncs, don't take
@sch and internally determine the scx_sched instance to abort. Currently,
it's always scx_root but once multiple scheduler support is in place, it
will be the scx_sched instance that invoked the kfunc. This simplifies
many callsites and defers scx_sched lookup until error is triggered.
- @sch is propagated to ops_cpu_valid() and ops_sanitize_err(). The CPU
validity conditions in ops_cpu_valid() are factored into __cpu_valid() to
implement kf_cpu_valid() which is the counterpart to scx_kf_exit/error().
- All users are converted. Most conversions are straightforward.
check_rq_for_timeouts() and scx_softlockup() are updated to use explicit
rcu_dereference*(scx_root) for safety as they may execute asynchronous to
the exit path. scx_tick() is also updated to use rcu_dereference(). While
not strictly necessary due to the preceding scx_enabled() test and IRQ
disabled context, this removes the subtlety at no noticeable cost.
No behavior changes intended.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
__scx_exit() is the base exit implementation and there are three wrappers on
top of it - scx_exit(), __scx_error() and scx_error(). This is more
confusing than helpful especially given that there are only a couple users
of scx_exit() and __scx_error(). To simplify the situation:
- Make __scx_exit() take va_list and rename it to scx_vexit(). This is to
ease implementing more complex extensions on top.
- Make scx_exit() a varargs wrapper around __scx_exit(). scx_exit() now
takes both @kind and @exit_code.
- Convert existing scx_exit() and __scx_error() users to use the new
scx_exit().
- scx_error() remains unchanged.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
In preparation of hierarchical scheduling support, make SCX_CALL_OP*() take
explicit @sch instead of assuming scx_root. As scx_root is still the only
scheduler instance, this patch doesn't make any functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
- Always cache scx_root into local variable sch before using.
- Don't use scx_root if cached sch is available.
- Wrap !sch test with unlikely().
- Pass @scx into scx_cgroup_init/exit().
No behavior changes intended.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Merge the mmc fixes for v6.15-rc[n] into the next branch, to allow them to
get tested together with the new mmc changes that are targeted for v6.16.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
It is not obvious that this functions checks capabilities. Rename it to
include '_can' like other capability helpers and reword it slightly.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20250501063325.7262-10-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Fix warnings when CONFIG_PM=y and CONFIG_PM_SLEEP is not set by
surrounding the 2 functions with #ifdef CONFIG_PM_SLEEP.
drivers/mmc/host/sdhci-esdhc-imx.c:1659:13: warning: 'sdhc_esdhc_tuning_restore' defined but not used [-Wunused-function]
1659 | static void sdhc_esdhc_tuning_restore(struct sdhci_host *host)
drivers/mmc/host/sdhci-esdhc-imx.c:1637:13: warning: 'sdhc_esdhc_tuning_save' defined but not used [-Wunused-function]
1637 | static void sdhc_esdhc_tuning_save(struct sdhci_host *host)
Fixes: 3d1eea493894 ("mmc: sdhci-esdhc-imx: Save tuning value when card stays powered in suspend")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Haibo Chen <haibo.chen@nxp.com>
Link: https://lore.kernel.org/r/20250424034610.441532-1-rdunlap@infradead.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Rewrite the textual description for the WonderMedia SDMMC controller
as YAML schema, and switch the filename to follow the compatible
string.
Signed-off-by: Alexey Charkov <alchark@gmail.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://lore.kernel.org/r/20250423-vt8500-sdmmc-binding-v2-1-ea4f17fd0638@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
RK3576's power domains have a peculiar design where the PD_NVM power
domain, of which the sdhci controller is a part, seemingly does not have
idempotent runtime disable/enable. The end effect is that if PD_NVM gets
turned off by the generic power domain logic because all the devices
depending on it are suspended, then the next time the sdhci device is
unsuspended, it'll hang the SoC as soon as it tries accessing the CQHCI
registers.
RK3576's UFS support needed a new dev_pm_genpd_rpm_always_on function
added to the generic power domains API to handle what appears to be a
similar hardware design.
Use this new function to ask for the same treatment in the sdhci
controller by giving rk3576 its own platform data with its own postinit
function. The benefit of doing this instead of marking the power domains
always on in the power domain core is that we only do this if we know
the platform we're running on actually uses the sdhci controller. For
others, keeping PD_NVM always on would be a waste, as they won't run
into this specific issue. The only other IP in PD_NVM that could be
affected is FSPI0. If it gets a mainline driver, it will probably want
to do the same thing.
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
Fixes: cfee1b5077 ("pmdomain: rockchip: Add support for RK3576 SoC")
Cc: <stable@vger.kernel.org> # v6.15+
Link: https://lore.kernel.org/r/20250423-rk3576-emmc-fix-v3-1-0bf80e29967f@collabora.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The pointer 'extern struct kmem_cache *idxd_desc_pool' and the macro
'#define IDXD_ALLOCATED_BATCH_SIZE 128U' were introduced in commit
bfe1d56091 ("dmaengine: idxd: Init and probe for Intel data
accelerators") but they were never used.
Signed-off-by: Eder Zulian <ezulian@redhat.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20250408173829.892317-1-ezulian@redhat.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
cavium-thunderx enables its PCI device with pcim_enable_device(). This,
implicitly, switches the function pci_request_regions() into managed
mode, where it becomes a devres function.
The PCI subsystem wants to remove this hybrid nature from its
interfaces. To do so, users of the aforementioned combination of
functions must be ported to non-hybrid functions.
Moreover, since both functions are already managed in this driver, the
calls to pci_release_regions() are unnecessary.
Remove the calls to pci_release_regions().
Replace the call to sometimes-managed pci_request_regions() with one to
the always-managed pcim_request_all_regions().
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://lore.kernel.org/r/20250417092742.27887-2-phasta@kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Add a compatible for the MediaTek Dimensity 1200 (MT6893) SoC.
All of the MMC/SD controllers in this chip are compatible with
the ones found in MT8183, but do also make use of an optional
crypto clock when enabling HW disk encryption.
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://lore.kernel.org/r/20250416120245.147951-1-angelogioacchino.delregno@collabora.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The sdhci IP of SG2044 is similar to it of SG2042. They
share the same clock and controller configuration.
Add compatible string for SG2044.
Signed-off-by: Inochi Amaoto <inochiama@gmail.com>
Reviewed-by: Chen Wang <unicorn_wang@outlook.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20250413223507.46480-8-inochiama@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This change add a condition for 'single' burst type selection.
Read AXI_LEN field from EMMC50_CFG2(AHB2AXI wrapper) register, if the
value is not 0, it means the HWIP is using AXI as AMBA bus, which do
not support 'single' burst type.
Suggested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Axe Yang <axe.yang@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20250411054134.31822-1-axe.yang@mediatek.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Current standard tuning has some limitations:
1. Standard tuning only try 40 times to find first pass window, but this
pass window maybe not the best pass window.
2. Sometimes there are two tuning pass windows and the gap between
those two windows may only have one cell. If tuning step > 1, the gap may
just be skipped and host assumes those two windows as a continuous
windows. This will cause a bad delay cell near the gap to be selected.
3. Standard tuning logic need to detect at least one success and failure
to pass the tuning. If all cells in the tuning window pass, the hardware
will not set the SDHCI_CTRL_TUNED_CLK bit, causing tuning failed.
4. Standard tuning logic only check the CRC, do not really compare the data
pattern. If data pins are connected incorrectly, standard will not detect
this kind of issue.
Switch to manual tuning to avoid those limitations
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-7-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Enhance manual tuning configuration reliability by adding tuning control
status checks per the i.MX Reference Manual recommendations.
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-6-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Expand the auto-tuning window width from 0 to 3 for manual tuning to
account for sampling point shifts caused by temperature change. This change
is based on hardware recommendation, providing enough margin for the
auto-tuning logic to locate valid sampling points.
When config the manual tuning final sample delay, need deduct the auto
tuning window width according to the IP logic.
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-5-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Expand the auto-tuning window width from 2 to 3 for standard tuning to
account for sampling point shifts caused by temperature change. This change
is based on hardware recommendation, providing 50% more margin for the
auto-tuning logic to locate valid sampling points.
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-4-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
During manual tuning, residual data in the async FIFO from previous
commands may impact with the tuning process. To ensure a clean state:
1. Set the RST_FIFO bit (SYS_CTRL[22]) to reset the async FIFO.
2. Poll the bit until self-cleared, confirming reset completion.
This hardening ensures the tuning command starts with a clean FIFO state,
improving the reliability of the manual tuning procedure.
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-3-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
According to the i.MX Reference Manual, the RSTT bit (SYS_CTRL[28]) is
designed to reset the tuning circuit. While the Reference Manual states
that clearing EXECUTE_TUNING bit from 1 to 0 in AUTOCMD12_ERR_STATUS
can also set RSTT, this mechanism only works when the original
EXECUTE_TUNING bit was 1. When the bit is already 0, the tuning circuit
reset will not be triggered.
This explicit reset approach strengthens the tuning reliability and
aligns with the Reference Manual recommendations.
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409075550.3413032-2-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Calculate data timeout value based on clock instead of using max value.
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com>
Reviewed-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20250409072604.3410459-1-ziniu.wang_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
mmc_can_* functions sometimes relate to the card and sometimes to the host.
Make it obvious by renaming this function to include 'host'.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20250401095847.29271-12-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
mmc_can_* functions sometimes relate to the card and sometimes to the host.
Make it obvious by renaming this function to include 'host'.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20250401095847.29271-11-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
mmc_can_* functions sometimes relate to the card and sometimes to the host.
Make it obvious by renaming this function to include 'card'. Also, convert to
proper bool type while we are here.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20250401095847.29271-10-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
