When CONFIG_OF is disabled, build testing on x86 runs into a couple of
objtool warnings from functions that unconditionally call panic() but
have no __noreturn annotation:
vmlinux.o: warning: objtool: exynos3250_cmu_isp_probe+0x17: samsung_cmu_register_one() is missing a __noreturn annotation
vmlinux.o: warning: objtool: exynos7885_cmu_probe+0x16: exynos_arm64_register_cmu() is missing a __noreturn annotation
vmlinux.o: warning: objtool: exynos850_cmu_probe+0x16: exynos_arm64_register_cmu() is missing a __noreturn annotation
vmlinux.o: warning: objtool: exynosautov9_cmu_probe+0x16: exynos_arm64_register_cmu() is missing a __noreturn annotation
The objtool analysis is correct, and this could be addressed by just
returning success whenever CONFIG_OF is disabled to let all that code
be eliminated, but since the driver is no use without CONFIG_OF,
just add that as a dependency. It will still get compile tested on
all architectures since CONFIG_OF is enabled in allmodconfig and most
randconfig builds.
Cc: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20230609081559.915867-1-arnd@kernel.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230312161512.2715500-25-u.kleine-koenig@pengutronix.de
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
of_device_get_match_data() function should not be used on the device
other than the one matched to the given driver, because it always returns
the match_data of the matched driver. In case of exynos-clkout driver,
the code matched the OF IDs on the PARENT device, so replacing it with
of_device_get_match_data() broke the driver.
This reverts commit 777aaf3d1d.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Fixes: 777aaf3d1d ("clk: samsung: exynos-clkout: Use of_device_get_match_data()")
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20221108213718.32076-1-m.szyprowski@samsung.com
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
"div4" DIVs which divide PLLs by 4 are actually dividing "div2" DIVs by
2 to achieve a by 4 division, thus their parents are the respective
"div2" DIVs. These DIVs were mistakenly set to have the PLLs as parents.
This leads to the kernel thinking "div4"s and everything under them run
at 2x the clock speed. Fix this.
Fixes: 45bd8166a1 ("clk: samsung: Add initial Exynos7885 clock driver")
Signed-off-by: David Virag <virag.david003@gmail.com>
Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Link: https://lore.kernel.org/r/20221013151341.151208-1-virag.david003@gmail.com
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
CMU_MFCMSCL clock domain provides clocks for MFC (Multi-Format Codec),
JPEG Codec and Scaler IP-cores. According to Exynos850 TRM, CMU_MFCMSCL
generates MFC, M2M, MCSC and JPEG clocks for BLK_MFCMSCL.
This patch adds next clocks:
- bus clocks in CMU_TOP for CMU_MFCMSCL
- all internal CMU_MFCMSCL clocks
- leaf clocks for MFCMSCL, TZPC (TrustZone Protection Controller),
JPEG codec, M2M (Memory-to-Memory), MCSC (Multi-Channel Scaler),
MFC (Multi-Format Codec), PPMU (Platform Performance Monitoring
Unit), SysMMU and SysReg
MFCMSCL related gate clocks in CMU_TOP were marked as CLK_IS_CRITICAL,
because:
1. All of those have to be enabled in order to read
/sys/kernel/debug/clk/clk_summary file
2. When some user driver (e.g. exynos-sysmmu) disables some derived
leaf clock, it can lead to CMU_TOP clocks disable, which then makes
the system hang. To prevent that, the CLK_IS_CRITICAL flag is used,
as CLK_IGNORE_UNUSED is not enough.
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220809113323.29965-8-semen.protsenko@linaro.org
CMU_IS clock domain provides clocks for IS IP-core (Image Signal
Processing Subsystem). According to Exynos850 TRM, CMU_IS generates
CSIS, IPP, ITP, VRA and GDC clocks for BLK_IS.
This patch adds next clocks:
- bus clocks in CMU_TOP needed for CMU_IS
- all internal CMU_IS clocks
- leaf clocks for IS IP-core, CSIS (Camera Serial Interface Slave),
D_TZPC (TrustZone Protection Controller), CSIS DMA, GDC (Geometric
Distortion Correction), IPP (Image Preprocessing Processing core),
ITP (Image Texture Processing core), MCSC (Multi-Channel Scaler),
VRA (Visual Recognition Accelerator), PPMU (Platform Performance
Monitoring Unit), SysMMU and SysReg
IS related gate clocks in CMU_TOP were marked as CLK_IS_CRITICAL,
because:
1. All of those have to be enabled in order to read
/sys/kernel/debug/clk/clk_summary file
2. When some user driver (e.g. exynos-sysmmu) disables some derived
leaf clock, it can lead to CMU_TOP clocks disable, which then makes
the system hang. To prevent that, the CLK_IS_CRITICAL flag is used,
as CLK_IGNORE_UNUSED is not enough.
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220809113323.29965-7-semen.protsenko@linaro.org
CMU_AUD clock domain provides clocks for ABOX IP-core (audio subsystem).
According to Exynos850 TRM, CMU_AUD generates Cortex-A32 clock, bus
clock and audio clocks for BLK_AUD.
This patch adds next clocks:
- bus clocks in CMU_TOP needed for CMU_AUD
- all internal CMU_AUD clocks
- leaf clocks for Cortex-A32, Speedy FM, UAIF0..UAIF6 (Unified Audio
Interface), CNT (counter), ABOX IP-core, ASB (Asynchronous Bridge),
DAP (Debug Access Port), I2S Codec MCLK, D_TZPC (TrustZone
Protection Controller), GPIO, PPMU (Platform Performance Monitoring
Unit), SysMMU, SysReg and WDT
ABOX clock was marked as CLK_IGNORE_UNUSED, as system hangs on boot
otherwise. Once ABOX driver is implemented, maybe it can be handled
there instead.
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220809113323.29965-6-semen.protsenko@linaro.org
TREX D Core and P core clocks seem to be related to the BTS (Bus Traffic
Shaper) inside the Exynos7885 SoC, and are needed for the SoC to
function correctly.
When clocks are cut from TREX D Core, the eMMC and the framebuffer stops
working properly. Other unknown things may stop working as well.
When clocks are cut from TREX P Core, the system locks up needing a hard
reset.
Add these clocks and mark them critical so that they are always on.
Signed-off-by: David Virag <virag.david003@gmail.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220601233743.56317-5-virag.david003@gmail.com
The Tesla FSD clock controller driver uses shared parts from Exynos
ARM64 clock drivers, so add proper dependency to fix COMPILE_TEST build
errors like:
/usr/bin/aarch64-linux-gnu-ld: drivers/clk/samsung/clk-fsd.o: in function `fsd_cmu_probe':
clk-fsd.c:(.init.text+0x9c): undefined reference to `exynos_arm64_register_cmu'
Reported-by: kernel test robot <lkp@intel.com>
Fixes: e3f3dc3810 ("clk: samsung: fsd: Add cmu_peric block clock information")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Alim Akhtar <alim.akhtar@samsung.com>
Link: https://lore.kernel.org/r/20220129173407.278591-1-krzysztof.kozlowski@canonical.com
This is an initial implementation adding basic clocks, such as UART,
USI, I2C, WDT, ect. and their parent clocks. It is heavily based on the
Exynos850 clock driver at 'drivers/clk/samsung/clk-exynos850.c' which
was made by Sam Protsenko, thus the copyright and author lines were
kept.
Bus clocks are enabled by default as well to avoid hangs while trying to
access CMU registers.
Only the parts of CMU_TOP needed for CMU_CORE and CMU_PERI, a bit of
CMU_CORE, and most of CMU_PERI is implemented as of now.
Signed-off-by: David Virag <virag.david003@gmail.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Link: https://lore.kernel.org/r/20211206153124.427102-7-virag.david003@gmail.com
pll1417x is used in Exynos7885 SoC for top-level integer PLLs.
It is similar enough to pll0822x that practically the same code can
handle both. The difference that's to be noted is that when defining a
pl1417x PLL, the "con" parameter of the PLL macro should be set to the
CON1 register instead of CON3, like this:
PLL(pll_1417x, CLK_FOUT_SHARED0_PLL, "fout_shared0_pll", "oscclk",
PLL_LOCKTIME_PLL_SHARED0, PLL_CON0_PLL_SHARED0,
NULL),
Signed-off-by: David Virag <virag.david003@gmail.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Link: https://lore.kernel.org/r/20211206153124.427102-6-virag.david003@gmail.com
