ARM: tegra: Don't enable unused PLLs on resume from suspend
Subsystems (3)
PATCH 1/1 ARM: tegra: Don't enable unused PLLs on resume from suspend
PLLC and PLLM are usually disabled on system suspend because all devices which use these PLLs are either suspended or switched away to other clock source. Don't enable unused PLLs on resume from suspend by keeping track of the enable-state of the PLLs across suspend-resume.
--- arch/arm/mach-tegra/sleep-tegra20.S | 38 ++++++++++-- arch/arm/mach-tegra/sleep-tegra30.S | 94 +++++++++++++++++++++++------ 2 files changed, 108 insertions(+), 24 deletions(-)
arch/arm/mach-tegra/sleep-tegra20.S +34 -4
@@ -43,11 +43,34 @@ #define APB_MISC_XM2CFGCPADCTRL2 0x8e4 #define APB_MISC_XM2CFGDPADCTRL2 0x8e8 -.macro pll_enable, rd, r_car_base, pll_base +#define PLLC_STORE_MASK (1 << 0) +#define PLLM_STORE_MASK (1 << 1) +#define PLLP_STORE_MASK (1 << 2) + +.macro test_pll_state, rd, test_mask + ldr \rd, tegra_pll_state + tst \rd, #\test_mask +.endm + +.macro store_pll_state, rd, tmp, r_car_base, pll_base, pll_mask + ldr \rd, [\r_car_base, #\pll_base] + tst \rd, #(1 << 30) + ldr \rd, tegra_pll_state + biceq \rd, \rd, #\pll_mask + orrne \rd, \rd, #\pll_mask + adr \tmp, tegra_pll_state + str \rd, [\tmp] +.endm + +.macro pll_enable, rd, r_car_base, pll_base, test_mask + test_pll_state \rd, \test_mask + beq 1f + ldr \rd, [\r_car_base, #\pll_base] tst \rd, #(1 << 30) orreq \rd, \rd, #(1 << 30) streq \rd, [\r_car_base, #\pll_base] +1: .endm .macro emc_device_mask, rd, base
@@ -177,9 +200,9 @@ ENTRY(tegra20_lp1_reset) str r1, [r0, #CLK_RESET_CCLK_DIVIDER] str r1, [r0, #CLK_RESET_SCLK_DIVIDER] - pll_enable r1, r0, CLK_RESET_PLLM_BASE - pll_enable r1, r0, CLK_RESET_PLLP_BASE - pll_enable r1, r0, CLK_RESET_PLLC_BASE + pll_enable r1, r0, CLK_RESET_PLLM_BASE, PLLM_STORE_MASK + pll_enable r1, r0, CLK_RESET_PLLP_BASE, PLLP_STORE_MASK + pll_enable r1, r0, CLK_RESET_PLLC_BASE, PLLC_STORE_MASK adr r2, tegra20_sdram_pad_address adr r4, tegra20_sdram_pad_save
@@ -270,6 +293,10 @@ tegra20_switch_cpu_to_clk32k: add r1, r1, #2 wait_until r1, r7, r9 + store_pll_state r0, r1, r5, CLK_RESET_PLLC_BASE, PLLC_STORE_MASK + store_pll_state r0, r1, r5, CLK_RESET_PLLM_BASE, PLLM_STORE_MASK + store_pll_state r0, r1, r5, CLK_RESET_PLLP_BASE, PLLP_STORE_MASK + /* disable PLLM, PLLP and PLLC */ ldr r0, [r5, #CLK_RESET_PLLM_BASE] bic r0, r0, #(1 << 30)
@@ -396,6 +423,9 @@ tegra20_sdram_pad_save: .long 0 .endr +tegra_pll_state: + .word 0x0 + .ltorg /* dummy symbol for end of IRAM */ .align L1_CACHE_SHIFT
arch/arm/mach-tegra/sleep-tegra30.S +74 -20
153 lines not on this page · raw message
Signature · 1 line
2.29.2
Quoted 5 lines · Dmitry Osipenko
PLLC and PLLM are usually disabled on system suspend because all devices which use these PLLs are either suspended or switched away to other clock source. Don't enable unused PLLs on resume from suspend by keeping track of the enable-state of the PLLs across suspend-resume.
Quoted 8 lines
Tested-by Peter Geis <[email protected]>Tested-by Nicolas Chauvet <[email protected]>Tested-by Matt Merhar <[email protected]>Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/mach-tegra/sleep-tegra20.S | 38 ++++++++++-- arch/arm/mach-tegra/sleep-tegra30.S | 94 +++++++++++++++++++++++------ 2 files changed, 108 insertions(+), 24 deletions(-)
For future patches with this many tested-bys, can you collect a bit of information about where these were tested? That'd be a good way to get a better understanding of the coverage and a good way to collect a bit of information about what devices are being used.
I've seen a pattern like this used sometimes:
So perhaps we can adopt this for these cases?
Applied, thanks.
Thierry
19.01.2021 20:24, Thierry Reding пишет:
Quoted 20 lines · Dmitry Osipenko
Quoted 5 lines
PLLC and PLLM are usually disabled on system suspend because all devices which use these PLLs are either suspended or switched away to other clock source. Don't enable unused PLLs on resume from suspend by keeping track of the enable-state of the PLLs across suspend-resume.
Quoted 8 lines
Tested-by Peter Geis <[email protected]>Tested-by Nicolas Chauvet <[email protected]>Tested-by Matt Merhar <[email protected]>Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/mach-tegra/sleep-tegra20.S | 38 ++++++++++-- arch/arm/mach-tegra/sleep-tegra30.S | 94 +++++++++++++++++++++++------ 2 files changed, 108 insertions(+), 24 deletions(-)For future patches with this many tested-bys, can you collect a bit of information about where these were tested? That'd be a good way to get a better understanding of the coverage and a good way to collect a bit of information about what devices are being used.
Quoted 7 lines
I've seen a pattern like this used sometimes:
Tested-by Name <email> # Board that was tested onSo perhaps we can adopt this for these cases?
Applied, thanks.
Okay, for this patch it should be: