* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver
@ 2018-08-29 21:09 Eddie James
2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James
` (5 more replies)
0 siblings, 6 replies; 21+ messages in thread
From: Eddie James @ 2018-08-29 21:09 UTC (permalink / raw)
To: linux-arm-kernel
The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs
can capture and compress video data from digital or analog sources. With
the Aspeed chip acting as a service processor, the Video Engine can
capture the host processor graphics output.
This series adds a V4L2 driver for the VE, providing a read() interface
only. The driver triggers the hardware to capture the host graphics output
and compress it to JPEG format.
Testing on an AST2500 determined that the videobuf/streaming/mmap interface
was significantly slower than the simple read() interface, so I have not
included the streaming part.
It's also possible to use an automatic mode for the VE such that
re-triggering the HW every frame isn't necessary. However this wasn't
reliable on the AST2400, and probably used more CPU anyway due to excessive
interrupts. It was approximately 15% faster.
The series also adds the necessary parent clock definitions to the Aspeed
clock driver, with both a mux and clock divider.
Eddie James (4):
clock: aspeed: Add VIDEO reset index definition
clock: aspeed: Setup video engine clocking
dt-bindings: media: Add Aspeed Video Engine binding documentation
media: platform: Add Aspeed Video Engine driver
.../devicetree/bindings/media/aspeed-video.txt | 23 +
drivers/clk/clk-aspeed.c | 41 +-
drivers/media/platform/Kconfig | 8 +
drivers/media/platform/Makefile | 1 +
drivers/media/platform/aspeed-video.c | 1307 ++++++++++++++++++++
include/dt-bindings/clock/aspeed-clock.h | 1 +
6 files changed, 1379 insertions(+), 2 deletions(-)
create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt
create mode 100644 drivers/media/platform/aspeed-video.c
--
1.8.3.1
^ permalink raw reply [flat|nested] 21+ messages in thread* [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James @ 2018-08-29 21:09 ` Eddie James 2018-08-31 17:56 ` Stephen Boyd 2018-09-04 13:36 ` Rob Herring 2018-08-29 21:09 ` [PATCH 2/4] clock: aspeed: Setup video engine clocking Eddie James ` (4 subsequent siblings) 5 siblings, 2 replies; 21+ messages in thread From: Eddie James @ 2018-08-29 21:09 UTC (permalink / raw) To: linux-arm-kernel Add an index into the array of resets kept in the Aspeed clock driver. This isn't a HW bit definition. Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> --- include/dt-bindings/clock/aspeed-clock.h | 1 + 1 file changed, 1 insertion(+) diff --git a/include/dt-bindings/clock/aspeed-clock.h b/include/dt-bindings/clock/aspeed-clock.h index f437386..15a9059 100644 --- a/include/dt-bindings/clock/aspeed-clock.h +++ b/include/dt-bindings/clock/aspeed-clock.h @@ -50,5 +50,6 @@ #define ASPEED_RESET_I2C 7 #define ASPEED_RESET_AHB 8 #define ASPEED_RESET_CRT1 9 +#define ASPEED_RESET_VIDEO 10 #endif -- 1.8.3.1 ^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition 2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James @ 2018-08-31 17:56 ` Stephen Boyd 2018-09-04 13:36 ` Rob Herring 1 sibling, 0 replies; 21+ messages in thread From: Stephen Boyd @ 2018-08-31 17:56 UTC (permalink / raw) To: linux-arm-kernel Quoting Eddie James (2018-08-29 14:09:30) > Add an index into the array of resets kept in the Aspeed clock driver. > This isn't a HW bit definition. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- Acked-by: Stephen Boyd <sboyd@kernel.org> ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition 2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James 2018-08-31 17:56 ` Stephen Boyd @ 2018-09-04 13:36 ` Rob Herring 1 sibling, 0 replies; 21+ messages in thread From: Rob Herring @ 2018-09-04 13:36 UTC (permalink / raw) To: linux-arm-kernel On Wed, 29 Aug 2018 16:09:30 -0500, Eddie James wrote: > Add an index into the array of resets kept in the Aspeed clock driver. > This isn't a HW bit definition. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- > include/dt-bindings/clock/aspeed-clock.h | 1 + > 1 file changed, 1 insertion(+) > Reviewed-by: Rob Herring <robh@kernel.org> ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 2/4] clock: aspeed: Setup video engine clocking 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James 2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James @ 2018-08-29 21:09 ` Eddie James 2018-08-31 17:56 ` Stephen Boyd 2018-08-31 21:33 ` Joel Stanley 2018-08-29 21:09 ` [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation Eddie James ` (3 subsequent siblings) 5 siblings, 2 replies; 21+ messages in thread From: Eddie James @ 2018-08-29 21:09 UTC (permalink / raw) To: linux-arm-kernel Add the video engine reset bit. Add eclk mux and clock divider table. Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> --- drivers/clk/clk-aspeed.c | 41 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 39 insertions(+), 2 deletions(-) diff --git a/drivers/clk/clk-aspeed.c b/drivers/clk/clk-aspeed.c index 5961367..f16ce7d 100644 --- a/drivers/clk/clk-aspeed.c +++ b/drivers/clk/clk-aspeed.c @@ -87,7 +87,7 @@ struct aspeed_clk_gate { /* TODO: ask Aspeed about the actual parent data */ static const struct aspeed_gate_data aspeed_gates[] = { /* clk rst name parent flags */ - [ASPEED_CLK_GATE_ECLK] = { 0, -1, "eclk-gate", "eclk", 0 }, /* Video Engine */ + [ASPEED_CLK_GATE_ECLK] = { 0, 6, "eclk-gate", "eclk", 0 }, /* Video Engine */ [ASPEED_CLK_GATE_GCLK] = { 1, 7, "gclk-gate", NULL, 0 }, /* 2D engine */ [ASPEED_CLK_GATE_MCLK] = { 2, -1, "mclk-gate", "mpll", CLK_IS_CRITICAL }, /* SDRAM */ [ASPEED_CLK_GATE_VCLK] = { 3, 6, "vclk-gate", NULL, 0 }, /* Video Capture */ @@ -113,6 +113,24 @@ struct aspeed_clk_gate { [ASPEED_CLK_GATE_LHCCLK] = { 28, -1, "lhclk-gate", "lhclk", 0 }, /* LPC master/LPC+ */ }; +static const char * const eclk_parent_names[] = { + "mpll", + "hpll", + "dpll", +}; + +static const struct clk_div_table ast2500_eclk_div_table[] = { + { 0x0, 2 }, + { 0x1, 2 }, + { 0x2, 3 }, + { 0x3, 4 }, + { 0x4, 5 }, + { 0x5, 6 }, + { 0x6, 7 }, + { 0x7, 8 }, + { 0 } +}; + static const struct clk_div_table ast2500_mac_div_table[] = { { 0x0, 4 }, /* Yep, really. Aspeed confirmed this is correct */ { 0x1, 4 }, @@ -192,18 +210,21 @@ static struct clk_hw *aspeed_ast2500_calc_pll(const char *name, u32 val) struct aspeed_clk_soc_data { const struct clk_div_table *div_table; + const struct clk_div_table *eclk_div_table; const struct clk_div_table *mac_div_table; struct clk_hw *(*calc_pll)(const char *name, u32 val); }; static const struct aspeed_clk_soc_data ast2500_data = { .div_table = ast2500_div_table, + .eclk_div_table = ast2500_eclk_div_table, .mac_div_table = ast2500_mac_div_table, .calc_pll = aspeed_ast2500_calc_pll, }; static const struct aspeed_clk_soc_data ast2400_data = { .div_table = ast2400_div_table, + .eclk_div_table = ast2400_div_table, .mac_div_table = ast2400_div_table, .calc_pll = aspeed_ast2400_calc_pll, }; @@ -317,6 +338,7 @@ struct aspeed_reset { [ASPEED_RESET_PECI] = 10, [ASPEED_RESET_I2C] = 2, [ASPEED_RESET_AHB] = 1, + [ASPEED_RESET_VIDEO] = 6, /* * SCUD4 resets start at an offset to separate them from @@ -522,6 +544,22 @@ static int aspeed_clk_probe(struct platform_device *pdev) return PTR_ERR(hw); aspeed_clk_data->hws[ASPEED_CLK_24M] = hw; + hw = clk_hw_register_mux(dev, "eclk-mux", eclk_parent_names, + ARRAY_SIZE(eclk_parent_names), 0, + scu_base + ASPEED_CLK_SELECTION, 2, 0x3, 0, + &aspeed_clk_lock); + if (IS_ERR(hw)) + return PTR_ERR(hw); + aspeed_clk_data->hws[ASPEED_CLK_ECLK_MUX] = hw; + + hw = clk_hw_register_divider_table(dev, "eclk", "eclk-mux", 0, + scu_base + ASPEED_CLK_SELECTION, 28, + 3, 0, soc_data->eclk_div_table, + &aspeed_clk_lock); + if (IS_ERR(hw)) + return PTR_ERR(hw); + aspeed_clk_data->hws[ASPEED_CLK_ECLK] = hw; + /* * TODO: There are a number of clocks that not included in this driver * as more information is required: @@ -531,7 +569,6 @@ static int aspeed_clk_probe(struct platform_device *pdev) * RGMII * RMII * UART[1..5] clock source mux - * Video Engine (ECLK) mux and clock divider */ for (i = 0; i < ARRAY_SIZE(aspeed_gates); i++) { -- 1.8.3.1 ^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 2/4] clock: aspeed: Setup video engine clocking 2018-08-29 21:09 ` [PATCH 2/4] clock: aspeed: Setup video engine clocking Eddie James @ 2018-08-31 17:56 ` Stephen Boyd 2018-08-31 21:33 ` Joel Stanley 1 sibling, 0 replies; 21+ messages in thread From: Stephen Boyd @ 2018-08-31 17:56 UTC (permalink / raw) To: linux-arm-kernel Quoting Eddie James (2018-08-29 14:09:31) > Add the video engine reset bit. Add eclk mux and clock divider table. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- Acked-by: Stephen Boyd <sboyd@kernel.org> ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 2/4] clock: aspeed: Setup video engine clocking 2018-08-29 21:09 ` [PATCH 2/4] clock: aspeed: Setup video engine clocking Eddie James 2018-08-31 17:56 ` Stephen Boyd @ 2018-08-31 21:33 ` Joel Stanley 1 sibling, 0 replies; 21+ messages in thread From: Joel Stanley @ 2018-08-31 21:33 UTC (permalink / raw) To: linux-arm-kernel On Wed, 29 Aug 2018 at 14:09, Eddie James <eajames@linux.vnet.ibm.com> wrote: > > Add the video engine reset bit. Add eclk mux and clock divider table. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> I'm travelling this week, so I'll be able to take a closer look next week. At a glance it looks okay. Cheers, Joel > --- > drivers/clk/clk-aspeed.c | 41 +++++++++++++++++++++++++++++++++++++++-- > 1 file changed, 39 insertions(+), 2 deletions(-) > > diff --git a/drivers/clk/clk-aspeed.c b/drivers/clk/clk-aspeed.c > index 5961367..f16ce7d 100644 > --- a/drivers/clk/clk-aspeed.c > +++ b/drivers/clk/clk-aspeed.c > @@ -87,7 +87,7 @@ struct aspeed_clk_gate { > /* TODO: ask Aspeed about the actual parent data */ > static const struct aspeed_gate_data aspeed_gates[] = { > /* clk rst name parent flags */ > - [ASPEED_CLK_GATE_ECLK] = { 0, -1, "eclk-gate", "eclk", 0 }, /* Video Engine */ > + [ASPEED_CLK_GATE_ECLK] = { 0, 6, "eclk-gate", "eclk", 0 }, /* Video Engine */ > [ASPEED_CLK_GATE_GCLK] = { 1, 7, "gclk-gate", NULL, 0 }, /* 2D engine */ > [ASPEED_CLK_GATE_MCLK] = { 2, -1, "mclk-gate", "mpll", CLK_IS_CRITICAL }, /* SDRAM */ > [ASPEED_CLK_GATE_VCLK] = { 3, 6, "vclk-gate", NULL, 0 }, /* Video Capture */ > @@ -113,6 +113,24 @@ struct aspeed_clk_gate { > [ASPEED_CLK_GATE_LHCCLK] = { 28, -1, "lhclk-gate", "lhclk", 0 }, /* LPC master/LPC+ */ > }; > > +static const char * const eclk_parent_names[] = { > + "mpll", > + "hpll", > + "dpll", > +}; > + > +static const struct clk_div_table ast2500_eclk_div_table[] = { > + { 0x0, 2 }, > + { 0x1, 2 }, > + { 0x2, 3 }, > + { 0x3, 4 }, > + { 0x4, 5 }, > + { 0x5, 6 }, > + { 0x6, 7 }, > + { 0x7, 8 }, > + { 0 } > +}; > + > static const struct clk_div_table ast2500_mac_div_table[] = { > { 0x0, 4 }, /* Yep, really. Aspeed confirmed this is correct */ > { 0x1, 4 }, > @@ -192,18 +210,21 @@ static struct clk_hw *aspeed_ast2500_calc_pll(const char *name, u32 val) > > struct aspeed_clk_soc_data { > const struct clk_div_table *div_table; > + const struct clk_div_table *eclk_div_table; > const struct clk_div_table *mac_div_table; > struct clk_hw *(*calc_pll)(const char *name, u32 val); > }; > > static const struct aspeed_clk_soc_data ast2500_data = { > .div_table = ast2500_div_table, > + .eclk_div_table = ast2500_eclk_div_table, > .mac_div_table = ast2500_mac_div_table, > .calc_pll = aspeed_ast2500_calc_pll, > }; > > static const struct aspeed_clk_soc_data ast2400_data = { > .div_table = ast2400_div_table, > + .eclk_div_table = ast2400_div_table, > .mac_div_table = ast2400_div_table, > .calc_pll = aspeed_ast2400_calc_pll, > }; > @@ -317,6 +338,7 @@ struct aspeed_reset { > [ASPEED_RESET_PECI] = 10, > [ASPEED_RESET_I2C] = 2, > [ASPEED_RESET_AHB] = 1, > + [ASPEED_RESET_VIDEO] = 6, > > /* > * SCUD4 resets start at an offset to separate them from > @@ -522,6 +544,22 @@ static int aspeed_clk_probe(struct platform_device *pdev) > return PTR_ERR(hw); > aspeed_clk_data->hws[ASPEED_CLK_24M] = hw; > > + hw = clk_hw_register_mux(dev, "eclk-mux", eclk_parent_names, > + ARRAY_SIZE(eclk_parent_names), 0, > + scu_base + ASPEED_CLK_SELECTION, 2, 0x3, 0, > + &aspeed_clk_lock); > + if (IS_ERR(hw)) > + return PTR_ERR(hw); > + aspeed_clk_data->hws[ASPEED_CLK_ECLK_MUX] = hw; > + > + hw = clk_hw_register_divider_table(dev, "eclk", "eclk-mux", 0, > + scu_base + ASPEED_CLK_SELECTION, 28, > + 3, 0, soc_data->eclk_div_table, > + &aspeed_clk_lock); > + if (IS_ERR(hw)) > + return PTR_ERR(hw); > + aspeed_clk_data->hws[ASPEED_CLK_ECLK] = hw; > + > /* > * TODO: There are a number of clocks that not included in this driver > * as more information is required: > @@ -531,7 +569,6 @@ static int aspeed_clk_probe(struct platform_device *pdev) > * RGMII > * RMII > * UART[1..5] clock source mux > - * Video Engine (ECLK) mux and clock divider > */ > > for (i = 0; i < ARRAY_SIZE(aspeed_gates); i++) { > -- > 1.8.3.1 > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James 2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James 2018-08-29 21:09 ` [PATCH 2/4] clock: aspeed: Setup video engine clocking Eddie James @ 2018-08-29 21:09 ` Eddie James 2018-09-04 0:50 ` Rob Herring 2018-08-29 21:09 ` [PATCH 4/4] media: platform: Add Aspeed Video Engine driver Eddie James ` (2 subsequent siblings) 5 siblings, 1 reply; 21+ messages in thread From: Eddie James @ 2018-08-29 21:09 UTC (permalink / raw) To: linux-arm-kernel Document the bindings. Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> --- .../devicetree/bindings/media/aspeed-video.txt | 23 ++++++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt diff --git a/Documentation/devicetree/bindings/media/aspeed-video.txt b/Documentation/devicetree/bindings/media/aspeed-video.txt new file mode 100644 index 0000000..58c056a --- /dev/null +++ b/Documentation/devicetree/bindings/media/aspeed-video.txt @@ -0,0 +1,23 @@ +* Device tree bindings for Aspeed Video Engine + +The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs can +capture and compress video data from digital or analog sources. + +Required properties: + - compatible: "aspeed,ast2400-video" or "aspeed,ast2500-video" + - reg: contains the offset and length of the VE memory region + - clocks: pointers to the the "vclk" and "eclk" of the syscon + - clock-names: "vclk-gate", "eclk-gate" + - resets: pointer to the VE reset of the syscon + - interrupts: the interrupt associated with the VE on this platform + +Example: + +video: video at 1e700000 { + compatible = "aspeed,ast2500-video"; + reg = <0x1e700000 0x20000>; + clocks = <&syscon ASPEED_CLK_GATE_VCLK>, <&syscon ASPEED_CLK_GATE_ECLK>; + clock-names = "vclk-gate", "eclk-gate"; + resets = <&syscon ASPEED_RESET_VIDEO>; + interrupts = <7>; +}; -- 1.8.3.1 ^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation 2018-08-29 21:09 ` [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation Eddie James @ 2018-09-04 0:50 ` Rob Herring 0 siblings, 0 replies; 21+ messages in thread From: Rob Herring @ 2018-09-04 0:50 UTC (permalink / raw) To: linux-arm-kernel On Wed, Aug 29, 2018 at 04:09:32PM -0500, Eddie James wrote: > Document the bindings. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- > .../devicetree/bindings/media/aspeed-video.txt | 23 ++++++++++++++++++++++ > 1 file changed, 23 insertions(+) > create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt > > diff --git a/Documentation/devicetree/bindings/media/aspeed-video.txt b/Documentation/devicetree/bindings/media/aspeed-video.txt > new file mode 100644 > index 0000000..58c056a > --- /dev/null > +++ b/Documentation/devicetree/bindings/media/aspeed-video.txt > @@ -0,0 +1,23 @@ > +* Device tree bindings for Aspeed Video Engine > + > +The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs can > +capture and compress video data from digital or analog sources. > + > +Required properties: > + - compatible: "aspeed,ast2400-video" or "aspeed,ast2500-video" Maybe "video-engine" or "ve" would be more specific than just "video". Use what is closest to the name of the block. > + - reg: contains the offset and length of the VE memory region > + - clocks: pointers to the the "vclk" and "eclk" of the syscon No pointers in DT. "Clock specifier" > + - clock-names: "vclk-gate", "eclk-gate" Is that the name at or in the module? The fact that it is gated is really outside the block and a property of the clock controller. > + - resets: pointer to the VE reset of the syscon > + - interrupts: the interrupt associated with the VE on this platform > + > +Example: > + > +video: video at 1e700000 { video-codec at ... > + compatible = "aspeed,ast2500-video"; > + reg = <0x1e700000 0x20000>; > + clocks = <&syscon ASPEED_CLK_GATE_VCLK>, <&syscon ASPEED_CLK_GATE_ECLK>; > + clock-names = "vclk-gate", "eclk-gate"; > + resets = <&syscon ASPEED_RESET_VIDEO>; > + interrupts = <7>; > +}; > -- > 1.8.3.1 > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 4/4] media: platform: Add Aspeed Video Engine driver 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James ` (2 preceding siblings ...) 2018-08-29 21:09 ` [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation Eddie James @ 2018-08-29 21:09 ` Eddie James 2018-08-30 0:52 ` Ezequiel Garcia 2018-09-03 11:40 ` Hans Verkuil 2018-08-31 17:56 ` [PATCH 0/4] " Stephen Boyd 2018-09-03 11:57 ` Hans Verkuil 5 siblings, 2 replies; 21+ messages in thread From: Eddie James @ 2018-08-29 21:09 UTC (permalink / raw) To: linux-arm-kernel The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs can capture and compress video data from digital or analog sources. With the Aspeed chip acting a service processor, the Video Engine can capture the host processor graphics output. Add a V4L2 driver to capture video data and compress it to JPEG images, making the data available through a standard read interface. Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> --- drivers/media/platform/Kconfig | 8 + drivers/media/platform/Makefile | 1 + drivers/media/platform/aspeed-video.c | 1307 +++++++++++++++++++++++++++++++++ 3 files changed, 1316 insertions(+) create mode 100644 drivers/media/platform/aspeed-video.c diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig index 94c1fe0..e599245 100644 --- a/drivers/media/platform/Kconfig +++ b/drivers/media/platform/Kconfig @@ -32,6 +32,14 @@ source "drivers/media/platform/davinci/Kconfig" source "drivers/media/platform/omap/Kconfig" +config VIDEO_ASPEED + tristate "Aspeed AST2400 and AST2500 Video Engine driver" + depends on VIDEO_V4L2 + help + Support for the Aspeed Video Engine (VE) embedded in the Aspeed + AST2400 and AST2500 SOCs. The VE can capture and compress video data + from digital or analog sources. + config VIDEO_SH_VOU tristate "SuperH VOU video output driver" depends on MEDIA_CAMERA_SUPPORT diff --git a/drivers/media/platform/Makefile b/drivers/media/platform/Makefile index 41322ab..205c33a 100644 --- a/drivers/media/platform/Makefile +++ b/drivers/media/platform/Makefile @@ -3,6 +3,7 @@ # Makefile for the video capture/playback device drivers. # +obj-$(CONFIG_VIDEO_ASPEED) += aspeed-video.o obj-$(CONFIG_VIDEO_CADENCE) += cadence/ obj-$(CONFIG_VIDEO_VIA_CAMERA) += via-camera.o obj-$(CONFIG_VIDEO_CAFE_CCIC) += marvell-ccic/ diff --git a/drivers/media/platform/aspeed-video.c b/drivers/media/platform/aspeed-video.c new file mode 100644 index 0000000..4f36a3f --- /dev/null +++ b/drivers/media/platform/aspeed-video.c @@ -0,0 +1,1307 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * ASPEED Video Engine Driver + * + * Copyright 2018 IBM Corp. + * + * Eddie James <eajames@linux.vnet.ibm.com> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/atomic.h> +#include <linux/bitfield.h> +#include <linux/clk.h> +#include <linux/delay.h> +#include <linux/device.h> +#include <linux/dma-mapping.h> +#include <linux/interrupt.h> +#include <linux/jiffies.h> +#include <linux/module.h> +#include <linux/mutex.h> +#include <linux/of.h> +#include <linux/of_irq.h> +#include <linux/of_reserved_mem.h> +#include <linux/platform_device.h> +#include <linux/reset.h> +#include <linux/sched.h> +#include <linux/spinlock.h> +#include <linux/string.h> +#include <linux/videodev2.h> +#include <linux/wait.h> +#include <linux/workqueue.h> +#include <media/v4l2-dev.h> +#include <media/v4l2-device.h> +#include <media/v4l2-ioctl.h> + +#define DEVICE_NAME "aspeed-video" + +#define ASPEED_VIDEO_JPEG_NUM_QUALITIES 12 +#define ASPEED_VIDEO_JPEG_HEADER_SIZE 10 +#define ASPEED_VIDEO_JPEG_QUANT_SIZE 116 +#define ASPEED_VIDEO_JPEG_DCT_SIZE 34 + +#define NUM_POLARITY_CHECKS 10 +#define INVALID_RESOLUTION_RETRIES 1 +#define INVALID_RESOLUTION_DELAY msecs_to_jiffies(250) +#define RESOLUTION_CHANGE_DELAY msecs_to_jiffies(500) +#define MODE_DETECT_TIMEOUT msecs_to_jiffies(500) +#define DIRECT_FETCH_THRESHOLD 0x0c0000 /* 1024 * 768, 32bpp */ + +#define VE_SRC_BUFFER_SIZE 0x900000 /* 1920 * 1200, 32bpp */ +#define VE_COMP_BUFFER_SIZE 0x100000 /* 128K packet * 8 packets */ +#define VE_JPEG_BUFFER_SIZE 0x006000 /* 512 * 12 * 4 */ + +#define VE_PROTECTION_KEY 0x000 +#define VE_PROTECTION_KEY_UNLOCK 0x1A038AA8 + +#define VE_SEQ_CTRL 0x004 +#define VE_SEQ_CTRL_TRIG_MODE_DET BIT(0) +#define VE_SEQ_CTRL_TRIG_CAPTURE BIT(1) +#define VE_SEQ_CTRL_FORCE_IDLE BIT(2) +#define VE_SEQ_CTRL_MULT_FRAME BIT(3) +#define VE_SEQ_CTRL_TRIG_COMP BIT(4) +#define VE_SEQ_CTRL_AUTO_COMP BIT(5) +#define VE_SEQ_CTRL_EN_WATCHDOG BIT(7) +#define VE_SEQ_CTRL_YUV420 BIT(10) +#define VE_SEQ_CTRL_COMP_FMT GENMASK(11, 10) +#define VE_SEQ_CTRL_HALT BIT(12) +#define VE_SEQ_CTRL_EN_WATCHDOG_COMP BIT(14) +#define VE_SEQ_CTRL_TRIG_JPG BIT(15) +#define VE_SEQ_CTRL_CAP_BUSY BIT(16) +#define VE_SEQ_CTRL_COMP_BUSY BIT(18) + +#ifdef CONFIG_MACH_ASPEED_G5 +#define VE_SEQ_CTRL_JPEG_MODE BIT(13) /* AST2500 */ +#else +#define VE_SEQ_CTRL_JPEG_MODE BIT(8) /* AST2400 */ +#endif /* CONFIG_MACH_ASPEED_G5 */ + +#define VE_CTRL 0x008 +#define VE_CTRL_HSYNC_POL BIT(0) +#define VE_CTRL_VSYNC_POL BIT(1) +#define VE_CTRL_SOURCE BIT(2) +#define VE_CTRL_INT_DE BIT(4) +#define VE_CTRL_DIRECT_FETCH BIT(5) +#define VE_CTRL_YUV BIT(6) +#define VE_CTRL_RGB BIT(7) +#define VE_CTRL_CAPTURE_FMT GENMASK(7, 6) +#define VE_CTRL_AUTO_OR_CURSOR BIT(8) +#define VE_CTRL_CLK_INVERSE BIT(11) +#define VE_CTRL_CLK_DELAY GENMASK(11, 9) +#define VE_CTRL_INTERLACE BIT(14) +#define VE_CTRL_HSYNC_POL_CTRL BIT(15) +#define VE_CTRL_FRC GENMASK(23, 16) + +#define VE_TGS_0 0x00c +#define VE_TGS_1 0x010 +#define VE_TGS_FIRST GENMASK(28, 16) +#define VE_TGS_LAST GENMASK(12, 0) + +#define VE_SCALING_FACTOR 0x014 +#define VE_SCALING_FILTER0 0x018 +#define VE_SCALING_FILTER1 0x01c +#define VE_SCALING_FILTER2 0x020 +#define VE_SCALING_FILTER3 0x024 + +#define VE_CAP_WINDOW 0x030 +#define VE_COMP_WINDOW 0x034 +#define VE_COMP_PROC_OFFSET 0x038 +#define VE_COMP_OFFSET 0x03c +#define VE_JPEG_ADDR 0x040 +#define VE_SRC0_ADDR 0x044 +#define VE_SRC_SCANLINE_OFFSET 0x048 +#define VE_SRC1_ADDR 0x04c +#define VE_COMP_ADDR 0x054 + +#define VE_STREAM_BUF_SIZE 0x058 +#define VE_STREAM_BUF_SIZE_N_PACKETS GENMASK(5, 3) +#define VE_STREAM_BUF_SIZE_P_SIZE GENMASK(2, 0) + +#define VE_COMP_CTRL 0x060 +#define VE_COMP_CTRL_VQ_DCT_ONLY BIT(0) +#define VE_COMP_CTRL_VQ_4COLOR BIT(1) +#define VE_COMP_CTRL_QUANTIZE BIT(2) +#define VE_COMP_CTRL_EN_BQ BIT(4) +#define VE_COMP_CTRL_EN_CRYPTO BIT(5) +#define VE_COMP_CTRL_DCT_CHR GENMASK(10, 6) +#define VE_COMP_CTRL_DCT_LUM GENMASK(15, 11) +#define VE_COMP_CTRL_EN_HQ BIT(16) +#define VE_COMP_CTRL_RSVD BIT(19) +#define VE_COMP_CTRL_ENCODE GENMASK(21, 20) +#define VE_COMP_CTRL_HQ_DCT_CHR GENMASK(26, 22) +#define VE_COMP_CTRL_HQ_DCT_LUM GENMASK(31, 27) + +#define VE_OFFSET_COMP_STREAM 0x078 + +#define VE_SRC_LR_EDGE_DET 0x090 +#define VE_SRC_LR_EDGE_DET_LEFT GENMASK(11, 0) +#define VE_SRC_LR_EDGE_DET_NO_V BIT(12) +#define VE_SRC_LR_EDGE_DET_NO_H BIT(13) +#define VE_SRC_LR_EDGE_DET_NO_DISP BIT(14) +#define VE_SRC_LR_EDGE_DET_NO_CLK BIT(15) +#define VE_SRC_LR_EDGE_DET_RT_SHF 16 +#define VE_SRC_LR_EDGE_DET_RT GENMASK(27, VE_SRC_LR_EDGE_DET_RT_SHF) +#define VE_SRC_LR_EDGE_DET_INTERLACE BIT(31) + +#define VE_SRC_TB_EDGE_DET 0x094 +#define VE_SRC_TB_EDGE_DET_TOP GENMASK(12, 0) +#define VE_SRC_TB_EDGE_DET_BOT_SHF 16 +#define VE_SRC_TB_EDGE_DET_BOT GENMASK(28, VE_SRC_TB_EDGE_DET_BOT_SHF) + +#define VE_MODE_DETECT_STATUS 0x098 +#define VE_MODE_DETECT_STATUS_VSYNC BIT(28) +#define VE_MODE_DETECT_STATUS_HSYNC BIT(29) + +#define VE_INTERRUPT_CTRL 0x304 +#define VE_INTERRUPT_STATUS 0x308 +#define VE_INTERRUPT_MODE_DETECT_WD BIT(0) +#define VE_INTERRUPT_CAPTURE_COMPLETE BIT(1) +#define VE_INTERRUPT_COMP_READY BIT(2) +#define VE_INTERRUPT_COMP_COMPLETE BIT(3) +#define VE_INTERRUPT_MODE_DETECT BIT(4) +#define VE_INTERRUPT_FRAME_COMPLETE BIT(5) +#define VE_INTERRUPT_DECODE_ERR BIT(6) +#define VE_INTERRUPT_HALT_READY BIT(8) +#define VE_INTERRUPT_HANG_WD BIT(9) +#define VE_INTERRUPT_STREAM_DESC BIT(10) +#define VE_INTERRUPT_VSYNC_DESC BIT(11) + +#define VE_MODE_DETECT 0x30c +#define VE_MEM_RESTRICT_START 0x310 +#define VE_MEM_RESTRICT_END 0x314 + +enum { + VIDEO_MODE_DETECT_DONE, + VIDEO_RES_CHANGE, + VIDEO_FRAME_AVAILABLE, + VIDEO_FRAME_TRIGGERED, +}; + +struct aspeed_video_addr { + dma_addr_t dma; + void *virt; +}; + +struct aspeed_video { + void __iomem *base; + struct clk *eclk; + struct clk *vclk; + struct reset_control *rst; + + struct device *dev; + struct v4l2_device v4l2_dev; + struct video_device vdev; + struct mutex video_lock; + + atomic_t clients; + wait_queue_head_t wait; + struct delayed_work res_work; + unsigned long flags; + + int frame_idx; + u32 frame_size; + + dma_addr_t max; + dma_addr_t min; + struct aspeed_video_addr srcs[2]; + struct aspeed_video_addr comp[2]; + struct aspeed_video_addr jpeg; + + int frame_rate; + int jpeg_quality; + struct v4l2_pix_format fmt; +}; + +#define to_aspeed_video(x) container_of((x), struct aspeed_video, v4l2_dev) + +static const u32 aspeed_video_jpeg_header[ASPEED_VIDEO_JPEG_HEADER_SIZE] = { + 0xE0FFD8FF, 0x464A1000, 0x01004649, 0x60000101, 0x00006000, 0x0F00FEFF, + 0x00002D05, 0x00000000, 0x00000000, 0x00DBFF00 +}; + +static const u32 aspeed_video_jpeg_quant[ASPEED_VIDEO_JPEG_QUANT_SIZE] = { + 0x081100C0, 0x00000000, 0x00110103, 0x03011102, 0xC4FF0111, 0x00001F00, + 0x01010501, 0x01010101, 0x00000000, 0x00000000, 0x04030201, 0x08070605, + 0xFF0B0A09, 0x10B500C4, 0x03010200, 0x03040203, 0x04040505, 0x7D010000, + 0x00030201, 0x12051104, 0x06413121, 0x07615113, 0x32147122, 0x08A19181, + 0xC1B14223, 0xF0D15215, 0x72623324, 0x160A0982, 0x1A191817, 0x28272625, + 0x35342A29, 0x39383736, 0x4544433A, 0x49484746, 0x5554534A, 0x59585756, + 0x6564635A, 0x69686766, 0x7574736A, 0x79787776, 0x8584837A, 0x89888786, + 0x9493928A, 0x98979695, 0xA3A29A99, 0xA7A6A5A4, 0xB2AAA9A8, 0xB6B5B4B3, + 0xBAB9B8B7, 0xC5C4C3C2, 0xC9C8C7C6, 0xD4D3D2CA, 0xD8D7D6D5, 0xE2E1DAD9, + 0xE6E5E4E3, 0xEAE9E8E7, 0xF4F3F2F1, 0xF8F7F6F5, 0xC4FFFAF9, 0x00011F00, + 0x01010103, 0x01010101, 0x00000101, 0x00000000, 0x04030201, 0x08070605, + 0xFF0B0A09, 0x11B500C4, 0x02010200, 0x04030404, 0x04040507, 0x77020100, + 0x03020100, 0x21050411, 0x41120631, 0x71610751, 0x81322213, 0x91421408, + 0x09C1B1A1, 0xF0523323, 0xD1726215, 0x3424160A, 0x17F125E1, 0x261A1918, + 0x2A292827, 0x38373635, 0x44433A39, 0x48474645, 0x54534A49, 0x58575655, + 0x64635A59, 0x68676665, 0x74736A69, 0x78777675, 0x83827A79, 0x87868584, + 0x928A8988, 0x96959493, 0x9A999897, 0xA5A4A3A2, 0xA9A8A7A6, 0xB4B3B2AA, + 0xB8B7B6B5, 0xC3C2BAB9, 0xC7C6C5C4, 0xD2CAC9C8, 0xD6D5D4D3, 0xDAD9D8D7, + 0xE5E4E3E2, 0xE9E8E7E6, 0xF4F3F2EA, 0xF8F7F6F5, 0xDAFFFAF9, 0x01030C00, + 0x03110200, 0x003F0011 +}; + +static const u32 aspeed_video_jpeg_dct[ASPEED_VIDEO_JPEG_NUM_QUALITIES] + [ASPEED_VIDEO_JPEG_DCT_SIZE] = { + { 0x0D140043, 0x0C0F110F, 0x11101114, 0x17141516, 0x1E20321E, + 0x3D1E1B1B, 0x32242E2B, 0x4B4C3F48, 0x44463F47, 0x61735A50, + 0x566C5550, 0x88644644, 0x7A766C65, 0x4D808280, 0x8C978D60, + 0x7E73967D, 0xDBFF7B80, 0x1F014300, 0x272D2121, 0x3030582D, + 0x697BB958, 0xB8B9B97B, 0xB9B8A6A6, 0xB9B9B9B9, 0xB9B9B9B9, + 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, + 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xFFB9B9B9 }, + { 0x0C110043, 0x0A0D0F0D, 0x0F0E0F11, 0x14111213, 0x1A1C2B1A, + 0x351A1818, 0x2B1F2826, 0x4142373F, 0x3C3D373E, 0x55644E46, + 0x4B5F4A46, 0x77573D3C, 0x6B675F58, 0x43707170, 0x7A847B54, + 0x6E64836D, 0xDBFF6C70, 0x1B014300, 0x22271D1D, 0x2A2A4C27, + 0x5B6BA04C, 0xA0A0A06B, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, + 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, + 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xFFA0A0A0 }, + { 0x090E0043, 0x090A0C0A, 0x0C0B0C0E, 0x110E0F10, 0x15172415, + 0x2C151313, 0x241A211F, 0x36372E34, 0x31322E33, 0x4653413A, + 0x3E4E3D3A, 0x62483231, 0x58564E49, 0x385D5E5D, 0x656D6645, + 0x5B536C5A, 0xDBFF595D, 0x16014300, 0x1C201818, 0x22223F20, + 0x4B58853F, 0x85858558, 0x85858585, 0x85858585, 0x85858585, + 0x85858585, 0x85858585, 0x85858585, 0x85858585, 0x85858585, + 0x85858585, 0x85858585, 0x85858585, 0xFF858585 }, + { 0x070B0043, 0x07080A08, 0x0A090A0B, 0x0D0B0C0C, 0x11121C11, + 0x23110F0F, 0x1C141A19, 0x2B2B2429, 0x27282428, 0x3842332E, + 0x313E302E, 0x4E392827, 0x46443E3A, 0x2C4A4A4A, 0x50565137, + 0x48425647, 0xDBFF474A, 0x12014300, 0x161A1313, 0x1C1C331A, + 0x3D486C33, 0x6C6C6C48, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, + 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, + 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0xFF6C6C6C }, + { 0x06090043, 0x05060706, 0x07070709, 0x0A09090A, 0x0D0E160D, + 0x1B0D0C0C, 0x16101413, 0x21221C20, 0x1E1F1C20, 0x2B332824, + 0x26302624, 0x3D2D1F1E, 0x3735302D, 0x22393A39, 0x3F443F2B, + 0x38334338, 0xDBFF3739, 0x0D014300, 0x11130E0E, 0x15152613, + 0x2D355026, 0x50505035, 0x50505050, 0x50505050, 0x50505050, + 0x50505050, 0x50505050, 0x50505050, 0x50505050, 0x50505050, + 0x50505050, 0x50505050, 0x50505050, 0xFF505050 }, + { 0x04060043, 0x03040504, 0x05040506, 0x07060606, 0x09090F09, + 0x12090808, 0x0F0A0D0D, 0x16161315, 0x14151315, 0x1D221B18, + 0x19201918, 0x281E1514, 0x2423201E, 0x17262726, 0x2A2D2A1C, + 0x25222D25, 0xDBFF2526, 0x09014300, 0x0B0D0A0A, 0x0E0E1A0D, + 0x1F25371A, 0x37373725, 0x37373737, 0x37373737, 0x37373737, + 0x37373737, 0x37373737, 0x37373737, 0x37373737, 0x37373737, + 0x37373737, 0x37373737, 0x37373737, 0xFF373737 }, + { 0x02030043, 0x01020202, 0x02020203, 0x03030303, 0x04040704, + 0x09040404, 0x07050606, 0x0B0B090A, 0x0A0A090A, 0x0E110D0C, + 0x0C100C0C, 0x140F0A0A, 0x1211100F, 0x0B131313, 0x1516150E, + 0x12111612, 0xDBFF1213, 0x04014300, 0x05060505, 0x07070D06, + 0x0F121B0D, 0x1B1B1B12, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, + 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, + 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0xFF1B1B1B }, + { 0x01020043, 0x01010101, 0x01010102, 0x02020202, 0x03030503, + 0x06030202, 0x05030404, 0x07070607, 0x06070607, 0x090B0908, + 0x080A0808, 0x0D0A0706, 0x0C0B0A0A, 0x070C0D0C, 0x0E0F0E09, + 0x0C0B0F0C, 0xDBFF0C0C, 0x03014300, 0x03040303, 0x04040804, + 0x0A0C1208, 0x1212120C, 0x12121212, 0x12121212, 0x12121212, + 0x12121212, 0x12121212, 0x12121212, 0x12121212, 0x12121212, + 0x12121212, 0x12121212, 0x12121212, 0xFF121212 }, + { 0x01020043, 0x01010101, 0x01010102, 0x02020202, 0x03030503, + 0x06030202, 0x05030404, 0x07070607, 0x06070607, 0x090B0908, + 0x080A0808, 0x0D0A0706, 0x0C0B0A0A, 0x070C0D0C, 0x0E0F0E09, + 0x0C0B0F0C, 0xDBFF0C0C, 0x02014300, 0x03030202, 0x04040703, + 0x080A0F07, 0x0F0F0F0A, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, + 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, + 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0xFF0F0F0F }, + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x02020302, + 0x04020202, 0x03020303, 0x05050405, 0x05050405, 0x07080606, + 0x06080606, 0x0A070505, 0x09080807, 0x05090909, 0x0A0B0A07, + 0x09080B09, 0xDBFF0909, 0x02014300, 0x02030202, 0x03030503, + 0x07080C05, 0x0C0C0C08, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, + 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, + 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0xFF0C0C0C }, + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x01010201, + 0x03010101, 0x02010202, 0x03030303, 0x03030303, 0x04050404, + 0x04050404, 0x06050303, 0x06050505, 0x03060606, 0x07070704, + 0x06050706, 0xDBFF0606, 0x01014300, 0x01020101, 0x02020402, + 0x05060904, 0x09090906, 0x09090909, 0x09090909, 0x09090909, + 0x09090909, 0x09090909, 0x09090909, 0x09090909, 0x09090909, + 0x09090909, 0x09090909, 0x09090909, 0xFF090909 }, + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x01010101, + 0x01010101, 0x01010101, 0x01010101, 0x01010101, 0x02020202, + 0x02020202, 0x03020101, 0x03020202, 0x01030303, 0x03030302, + 0x03020303, 0xDBFF0403, 0x01014300, 0x01010101, 0x01010201, + 0x03040602, 0x06060604, 0x06060606, 0x06060606, 0x06060606, + 0x06060606, 0x06060606, 0x06060606, 0x06060606, 0x06060606, + 0x06060606, 0x06060606, 0x06060606, 0xFF060606 } +}; + +static void aspeed_video_init_jpeg_table(u32 *table, bool yuv420) +{ + int i; + unsigned int base; + + for (i = 0; i < ASPEED_VIDEO_JPEG_NUM_QUALITIES; i++) { + int j; + + base = 256 * i; /* AST HW requires this header spacing */ + + for (j = 0; j < ASPEED_VIDEO_JPEG_HEADER_SIZE; j++) + table[base + j] = + le32_to_cpu(aspeed_video_jpeg_header[j]); + + base += ASPEED_VIDEO_JPEG_HEADER_SIZE; + for (j = 0; j < ASPEED_VIDEO_JPEG_DCT_SIZE; j++) + table[base + j] = + le32_to_cpu(aspeed_video_jpeg_dct[i][j]); + + base += ASPEED_VIDEO_JPEG_DCT_SIZE; + for (j = 0; j < ASPEED_VIDEO_JPEG_QUANT_SIZE; j++) + table[base + j] = + le32_to_cpu(aspeed_video_jpeg_quant[j]); + + if (yuv420) + table[base + 2] = le32_to_cpu(0x00220103); + } +} + +static void aspeed_video_update(struct aspeed_video *video, u32 reg, + unsigned long mask, u32 bits) +{ + u32 t = readl(video->base + reg); + u32 before = t; + + t &= mask; + t |= bits; + writel(t, video->base + reg); + dev_dbg(video->dev, "update %03x[%08x -> %08x]\n", reg, before, + readl(video->base + reg)); +} + +static u32 aspeed_video_read(struct aspeed_video *video, u32 reg) +{ + u32 t = readl(video->base + reg); + + dev_dbg(video->dev, "read %03x[%08x]\n", reg, t); + return t; +} + +static void aspeed_video_write(struct aspeed_video *video, u32 reg, u32 val) +{ + writel(val, video->base + reg); + dev_dbg(video->dev, "write %03x[%08x]\n", reg, + readl(video->base + reg)); +} + +static bool aspeed_video_engine_busy(struct aspeed_video *video) +{ + u32 seq_ctrl = aspeed_video_read(video, VE_SEQ_CTRL); + + if (!(seq_ctrl & VE_SEQ_CTRL_COMP_BUSY) || + !(seq_ctrl & VE_SEQ_CTRL_CAP_BUSY)) { + dev_info(video->dev, "video engine busy\n"); + return true; + } + + return false; +} + +static int aspeed_video_start_frame(struct aspeed_video *video) +{ + if (aspeed_video_engine_busy(video)) + return -EBUSY; + + video->frame_idx = (video->frame_idx + 1) % 2; + + aspeed_video_write(video, VE_COMP_PROC_OFFSET, 0); + aspeed_video_write(video, VE_COMP_OFFSET, 0); + aspeed_video_write(video, VE_COMP_ADDR, + video->comp[video->frame_idx].dma); + + set_bit(VIDEO_FRAME_TRIGGERED, &video->flags); + + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, + VE_INTERRUPT_COMP_COMPLETE | + VE_INTERRUPT_CAPTURE_COMPLETE); + + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, + VE_SEQ_CTRL_TRIG_CAPTURE | VE_SEQ_CTRL_TRIG_COMP); + + return 0; +} + +static void aspeed_video_start_mode_detect(struct aspeed_video *video) +{ + /* Enable mode detect interrupts */ + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, + VE_INTERRUPT_MODE_DETECT); + + /* Trigger mode detect */ + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, + VE_SEQ_CTRL_TRIG_MODE_DET); +} + +static void aspeed_video_disable_mode_detect(struct aspeed_video *video) +{ + /* Disable mode detect interrupts */ + aspeed_video_update(video, VE_INTERRUPT_CTRL, + ~VE_INTERRUPT_MODE_DETECT, 0); + + /* Disable mode detect */ + aspeed_video_update(video, VE_SEQ_CTRL, ~VE_SEQ_CTRL_TRIG_MODE_DET, 0); +} + +static void aspeed_video_off(struct aspeed_video *video) +{ + /* Reset the engine */ + reset_control_assert(video->rst); + udelay(100); + reset_control_deassert(video->rst); + + /* Turn off the relevant clocks */ + clk_disable_unprepare(video->vclk); + clk_disable_unprepare(video->eclk); +} + +static void aspeed_video_on(struct aspeed_video *video) +{ + /* Turn on the relevant clocks */ + clk_prepare_enable(video->eclk); + clk_prepare_enable(video->vclk); + + /* Reset the engine */ + reset_control_assert(video->rst); + udelay(100); + reset_control_deassert(video->rst); +} + +static irqreturn_t aspeed_video_irq(int irq, void *arg) +{ + struct aspeed_video *video = arg; + u32 sts = aspeed_video_read(video, VE_INTERRUPT_STATUS); + + if (atomic_read(&video->clients) == 0) { + dev_info(video->dev, "irq with no client; disabling irqs\n"); + + aspeed_video_write(video, VE_INTERRUPT_CTRL, 0); + aspeed_video_write(video, VE_INTERRUPT_STATUS, 0xFFFFFFFF); + return IRQ_HANDLED; + } + + /* Resolution changed; reset entire engine and reinitialize */ + if (sts & VE_INTERRUPT_MODE_DETECT_WD) { + dev_info(video->dev, "resolution changed; resetting\n"); + set_bit(VIDEO_RES_CHANGE, &video->flags); + + aspeed_video_off(video); + + schedule_delayed_work(&video->res_work, + RESOLUTION_CHANGE_DELAY); + return IRQ_HANDLED; + } + + if (sts & VE_INTERRUPT_MODE_DETECT) { + aspeed_video_update(video, VE_INTERRUPT_CTRL, + ~VE_INTERRUPT_MODE_DETECT, 0); + aspeed_video_write(video, VE_INTERRUPT_STATUS, + VE_INTERRUPT_MODE_DETECT); + + set_bit(VIDEO_MODE_DETECT_DONE, &video->flags); + wake_up_interruptible_all(&video->wait); + } + + if ((sts & VE_INTERRUPT_COMP_COMPLETE) && + (sts & VE_INTERRUPT_CAPTURE_COMPLETE)) { + video->frame_size = aspeed_video_read(video, + VE_OFFSET_COMP_STREAM); + + aspeed_video_update(video, VE_INTERRUPT_CTRL, + ~(VE_INTERRUPT_COMP_COMPLETE | + VE_INTERRUPT_CAPTURE_COMPLETE), 0); + aspeed_video_write(video, VE_INTERRUPT_STATUS, + VE_INTERRUPT_COMP_COMPLETE | + VE_INTERRUPT_CAPTURE_COMPLETE); + aspeed_video_update(video, VE_SEQ_CTRL, + ~(VE_SEQ_CTRL_TRIG_CAPTURE | + VE_SEQ_CTRL_FORCE_IDLE | + VE_SEQ_CTRL_TRIG_COMP), 0); + + set_bit(VIDEO_FRAME_AVAILABLE, &video->flags); + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); + wake_up_interruptible_all(&video->wait); + } + + return IRQ_HANDLED; +} + +static void aspeed_video_check_polarity(struct aspeed_video *video) +{ + int i; + int hsync_counter = 0; + int vsync_counter = 0; + u32 sts; + + for (i = 0; i < NUM_POLARITY_CHECKS; ++i) { + sts = aspeed_video_read(video, VE_MODE_DETECT_STATUS); + if (sts & VE_MODE_DETECT_STATUS_VSYNC) + vsync_counter--; + else + vsync_counter++; + + if (sts & VE_MODE_DETECT_STATUS_HSYNC) + hsync_counter--; + else + hsync_counter++; + } + + if (hsync_counter < 0 || vsync_counter < 0) { + u32 ctrl; + + if (hsync_counter < 0) + ctrl = VE_CTRL_HSYNC_POL; + + if (vsync_counter < 0) + ctrl = VE_CTRL_VSYNC_POL; + + aspeed_video_update(video, VE_CTRL, 0xFFFFFFFF, ctrl); + } +} + +#define res_check(v) test_and_clear_bit(VIDEO_MODE_DETECT_DONE, &(v)->flags) + +static int aspeed_video_get_resolution(struct aspeed_video *video) +{ + bool invalid_resolution = true; + int rc; + int tries = 0; + unsigned int bottom; + unsigned int left; + unsigned int right; + unsigned int top; + u32 src_lr_edge; + u32 src_tb_edge; + + video->fmt.width = 0; + video->fmt.height = 0; + + do { + if (tries) { + set_current_state(TASK_INTERRUPTIBLE); + if (schedule_timeout(INVALID_RESOLUTION_DELAY)) + return -EINTR; + } + + aspeed_video_start_mode_detect(video); + + rc = wait_event_interruptible_timeout(video->wait, + res_check(video), + MODE_DETECT_TIMEOUT); + if (!rc) { + dev_err(video->dev, "timed out on 1st mode detect\n"); + aspeed_video_disable_mode_detect(video); + return -ETIME; + } + + /* Disable mode detect in order to re-trigger */ + aspeed_video_update(video, VE_SEQ_CTRL, + ~VE_SEQ_CTRL_TRIG_MODE_DET, 0); + + aspeed_video_check_polarity(video); + + aspeed_video_start_mode_detect(video); + + rc = wait_event_interruptible_timeout(video->wait, + res_check(video), + MODE_DETECT_TIMEOUT); + if (!rc) { + dev_err(video->dev, "timed out on 2nd mode detect\n"); + aspeed_video_disable_mode_detect(video); + return -ETIME; + } + + src_lr_edge = aspeed_video_read(video, VE_SRC_LR_EDGE_DET); + src_tb_edge = aspeed_video_read(video, VE_SRC_TB_EDGE_DET); + + bottom = (src_tb_edge & VE_SRC_TB_EDGE_DET_BOT) >> + VE_SRC_TB_EDGE_DET_BOT_SHF; + top = src_tb_edge & VE_SRC_TB_EDGE_DET_TOP; + if (top > bottom) + continue; + + right = (src_lr_edge & VE_SRC_LR_EDGE_DET_RT) >> + VE_SRC_LR_EDGE_DET_RT_SHF; + left = src_lr_edge & VE_SRC_LR_EDGE_DET_LEFT; + if (left > right) + continue; + + invalid_resolution = false; + } while (invalid_resolution && (tries++ < INVALID_RESOLUTION_RETRIES)); + + if (invalid_resolution) { + dev_err(video->dev, "invalid resolution detected\n"); + return -EMSGSIZE; + } + + video->fmt.height = (bottom - top) + 1; + video->fmt.width = (right - left) + 1; + + /* Don't use direct mode below 1024 x 768 (irqs don't fire) */ + if (video->fmt.height * video->fmt.width < DIRECT_FETCH_THRESHOLD) { + aspeed_video_write(video, VE_TGS_0, + FIELD_PREP(VE_TGS_FIRST, left - 1) | + FIELD_PREP(VE_TGS_LAST, right)); + aspeed_video_write(video, VE_TGS_1, + FIELD_PREP(VE_TGS_FIRST, top) | + FIELD_PREP(VE_TGS_LAST, bottom + 1)); + aspeed_video_update(video, VE_CTRL, + ~VE_CTRL_DIRECT_FETCH, VE_CTRL_INT_DE); + } + + aspeed_video_write(video, VE_CAP_WINDOW, + video->fmt.width << 16 | video->fmt.height); + aspeed_video_write(video, VE_COMP_WINDOW, + video->fmt.width << 16 | video->fmt.height); + aspeed_video_write(video, VE_SRC_SCANLINE_OFFSET, + video->fmt.width * 4); + + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, + VE_INTERRUPT_MODE_DETECT_WD); + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, + VE_SEQ_CTRL_EN_WATCHDOG); + + dev_dbg(video->dev, "got resolution[%dx%d]\n", video->fmt.width, + video->fmt.height); + + return 0; +} + +static void aspeed_video_init_regs(struct aspeed_video *video) +{ + u32 comp_ctrl = VE_COMP_CTRL_RSVD | + FIELD_PREP(VE_COMP_CTRL_DCT_LUM, video->jpeg_quality) | + FIELD_PREP(VE_COMP_CTRL_DCT_CHR, video->jpeg_quality | 0x10); + u32 ctrl = VE_CTRL_DIRECT_FETCH | VE_CTRL_AUTO_OR_CURSOR; + u32 seq_ctrl = VE_SEQ_CTRL_AUTO_COMP | VE_SEQ_CTRL_JPEG_MODE; + + if (video->frame_rate) + ctrl |= FIELD_PREP(VE_CTRL_FRC, video->frame_rate); + + if (video->fmt.pixelformat == V4L2_PIX_FMT_YUV420) + seq_ctrl |= VE_SEQ_CTRL_YUV420; + + /* Unlock VE registers */ + aspeed_video_write(video, VE_PROTECTION_KEY, VE_PROTECTION_KEY_UNLOCK); + + /* Disable interrupts */ + aspeed_video_write(video, VE_INTERRUPT_CTRL, 0); + aspeed_video_write(video, VE_INTERRUPT_STATUS, 0xFFFFFFFF); + + /* Clear the offset */ + aspeed_video_write(video, VE_COMP_PROC_OFFSET, 0); + aspeed_video_write(video, VE_COMP_OFFSET, 0); + + /* Set memory restrictions */ + aspeed_video_write(video, VE_MEM_RESTRICT_START, video->min); + aspeed_video_write(video, VE_MEM_RESTRICT_END, video->max); + + /* Set buffer addresses */ + aspeed_video_write(video, VE_SRC0_ADDR, video->srcs[0].dma); + aspeed_video_write(video, VE_SRC1_ADDR, video->srcs[1].dma); + aspeed_video_write(video, VE_COMP_ADDR, video->comp[0].dma); + aspeed_video_write(video, VE_JPEG_ADDR, video->jpeg.dma); + + /* Set control registers */ + aspeed_video_write(video, VE_SEQ_CTRL, seq_ctrl); + aspeed_video_write(video, VE_CTRL, ctrl); + aspeed_video_write(video, VE_COMP_CTRL, comp_ctrl); + + /* Compression buffer size; 128K packet * 8 packets */ + aspeed_video_write(video, VE_STREAM_BUF_SIZE, 0xf); + + /* Don't downscale */ + aspeed_video_write(video, VE_SCALING_FACTOR, 0x10001000); + aspeed_video_write(video, VE_SCALING_FILTER0, 0x00200000); + aspeed_video_write(video, VE_SCALING_FILTER1, 0x00200000); + aspeed_video_write(video, VE_SCALING_FILTER2, 0x00200000); + aspeed_video_write(video, VE_SCALING_FILTER3, 0x00200000); + + /* Set mode detection defaults */ + aspeed_video_write(video, VE_MODE_DETECT, 0x22666500); +} + +static int aspeed_video_allocate_cma(struct aspeed_video *video) +{ + video->srcs[0].virt = dma_alloc_coherent(video->dev, + VE_SRC_BUFFER_SIZE, + &video->srcs[0].dma, + GFP_KERNEL); + if (!video->srcs[0].virt) { + dev_err(video->dev, + "Failed to allocate source buffer 0, size[%x]\n", + VE_SRC_BUFFER_SIZE); + goto err; + } + + video->srcs[1].virt = dma_alloc_coherent(video->dev, + VE_SRC_BUFFER_SIZE, + &video->srcs[1].dma, + GFP_KERNEL); + if (!video->srcs[1].virt) { + dev_err(video->dev, + "Failed to allocate source buffer 1, size[%x]\n", + VE_SRC_BUFFER_SIZE); + goto free_src0; + } + + video->comp[0].virt = dma_alloc_coherent(video->dev, + VE_COMP_BUFFER_SIZE, + &video->comp[0].dma, + GFP_KERNEL); + if (!video->comp[0].virt) { + dev_err(video->dev, + "Failed to allocate compression buffer 0, size[%x]\n", + VE_COMP_BUFFER_SIZE); + goto free_src1; + } + + video->comp[1].virt = dma_alloc_coherent(video->dev, + VE_COMP_BUFFER_SIZE, + &video->comp[1].dma, + GFP_KERNEL); + if (!video->comp[0].virt) { + dev_err(video->dev, + "Failed to allocate compression buffer 1, size[%x]\n", + VE_COMP_BUFFER_SIZE); + goto free_comp0; + } + + video->jpeg.virt = dma_alloc_coherent(video->dev, VE_JPEG_BUFFER_SIZE, + &video->jpeg.dma, GFP_KERNEL); + if (!video->jpeg.virt) { + dev_err(video->dev, + "Failed to allocate JPEG buffer, size[%x]\n", + VE_JPEG_BUFFER_SIZE); + goto free_comp1; + } + + if (video->fmt.pixelformat == V4L2_PIX_FMT_YUV420) + aspeed_video_init_jpeg_table(video->jpeg.virt, true); + else + aspeed_video_init_jpeg_table(video->jpeg.virt, false); + + /* + * Calculate the memory restrictions. Don't consider the JPEG header + * buffer since HW doesn't need to write to it. + */ + video->max = max(video->srcs[0].dma + VE_SRC_BUFFER_SIZE, + video->srcs[1].dma + VE_SRC_BUFFER_SIZE); + video->max = max(video->max, video->comp[0].dma + VE_COMP_BUFFER_SIZE); + video->max = max(video->max, video->comp[1].dma + VE_COMP_BUFFER_SIZE); + + video->min = min(video->srcs[0].dma, video->srcs[1].dma); + video->min = min(video->min, video->comp[0].dma); + video->min = min(video->min, video->comp[1].dma); + + return 0; + +free_comp1: + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[1].virt, + video->comp[1].dma); + video->comp[1].dma = 0ULL; + video->comp[1].virt = NULL; + +free_comp0: + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[0].virt, + video->comp[0].dma); + video->comp[0].dma = 0ULL; + video->comp[0].virt = NULL; + +free_src1: + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[1].virt, + video->srcs[1].dma); + video->srcs[1].dma = 0ULL; + video->srcs[1].virt = NULL; + +free_src0: + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[0].virt, + video->srcs[0].dma); + video->srcs[0].dma = 0ULL; + video->srcs[0].virt = NULL; +err: + return -ENOMEM; +} + +static void aspeed_video_free_cma(struct aspeed_video *video) +{ + dma_free_coherent(video->dev, VE_JPEG_BUFFER_SIZE, video->jpeg.virt, + video->jpeg.dma); + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[1].virt, + video->comp[1].dma); + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[0].virt, + video->comp[0].dma); + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[1].virt, + video->srcs[1].dma); + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[0].virt, + video->srcs[0].dma); + + video->srcs[0].dma = 0ULL; + video->srcs[0].virt = NULL; + video->srcs[1].dma = 0ULL; + video->srcs[1].virt = NULL; + video->comp[0].dma = 0ULL; + video->comp[0].virt = NULL; + video->comp[1].dma = 0ULL; + video->comp[1].virt = NULL; + video->jpeg.dma = 0ULL; + video->jpeg.virt = NULL; +} + +static int aspeed_video_start(struct aspeed_video *video) +{ + int rc = aspeed_video_allocate_cma(video); + + if (rc) + return rc; + + aspeed_video_on(video); + + aspeed_video_init_regs(video); + + rc = aspeed_video_get_resolution(video); + if (rc) + aspeed_video_free_cma(video); + + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); + + return rc; +} + +static void aspeed_video_stop(struct aspeed_video *video) +{ + cancel_delayed_work_sync(&video->res_work); + + aspeed_video_off(video); + + aspeed_video_free_cma(video); + + clear_bit(VIDEO_FRAME_AVAILABLE, &video->flags); +} + +static int aspeed_video_querycap(struct file *file, void *fh, + struct v4l2_capability *cap) +{ + struct aspeed_video *video = video_drvdata(file); + + strncpy(cap->driver, DEVICE_NAME, sizeof(cap->driver)); + cap->capabilities = video->vdev.device_caps | V4L2_CAP_DEVICE_CAPS; + + return 0; +} + +static int aspeed_video_get_format(struct file *file, void *fh, + struct v4l2_format *f) +{ + int rc; + struct aspeed_video *video = video_drvdata(file); + + if (test_bit(VIDEO_RES_CHANGE, &video->flags)) { + if (file->f_flags & O_NONBLOCK) + return -EAGAIN; + + rc = wait_event_interruptible(video->wait, + !test_bit(VIDEO_RES_CHANGE, + &video->flags)); + if (rc) + return -EINTR; + } + + f->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; + f->fmt.pix = video->fmt; + + return 0; +} + +static int aspeed_video_set_format(struct file *file, void *fh, + struct v4l2_format *f) +{ + struct aspeed_video *video = video_drvdata(file); + + if (f->fmt.pix.pixelformat == video->fmt.pixelformat) + return 0; + + if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV444) { + video->fmt.pixelformat = V4L2_PIX_FMT_YUV444; + aspeed_video_init_jpeg_table(video->jpeg.virt, false); + aspeed_video_update(video, VE_SEQ_CTRL, ~VE_SEQ_CTRL_YUV420, + 0); + } else if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV420) { + video->fmt.pixelformat = V4L2_PIX_FMT_YUV420; + aspeed_video_init_jpeg_table(video->jpeg.virt, true); + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, + VE_SEQ_CTRL_YUV420); + } else { + return -EINVAL; + } + + return 0; +} + +static int aspeed_video_get_jpegcomp(struct file *file, void *fh, + struct v4l2_jpegcompression *a) +{ + struct aspeed_video *video = video_drvdata(file); + + a->quality = video->jpeg_quality; + + return 0; +} + +static int aspeed_video_set_jpegcomp(struct file *file, void *fh, + const struct v4l2_jpegcompression *a) +{ + u32 comp_ctrl; + struct aspeed_video *video = video_drvdata(file); + + if (a->quality < 0 || a->quality > 11) + return -EINVAL; + + video->jpeg_quality = a->quality; + comp_ctrl = FIELD_PREP(VE_COMP_CTRL_DCT_LUM, video->jpeg_quality) | + FIELD_PREP(VE_COMP_CTRL_DCT_CHR, video->jpeg_quality | 0x10); + + aspeed_video_update(video, VE_COMP_CTRL, + ~(VE_COMP_CTRL_DCT_LUM | VE_COMP_CTRL_DCT_CHR), + comp_ctrl); + + return 0; +} + +static int aspeed_video_get_parm(struct file *file, void *fh, + struct v4l2_streamparm *a) +{ + struct aspeed_video *video = video_drvdata(file); + + a->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; + a->parm.capture.timeperframe.numerator = 1; + a->parm.capture.timeperframe.denominator = video->frame_rate; + + return 0; +} + +static int aspeed_video_set_parm(struct file *file, void *fh, + struct v4l2_streamparm *a) +{ + int frame_rate; + struct aspeed_video *video = video_drvdata(file); + + if (a->type != V4L2_BUF_TYPE_VIDEO_CAPTURE) + return -EINVAL; + + frame_rate = a->parm.capture.timeperframe.denominator / + a->parm.capture.timeperframe.numerator; + + if (frame_rate < 0 || frame_rate > 60) + return -EINVAL; + + if (video->frame_rate != frame_rate) { + video->frame_rate = frame_rate; + aspeed_video_update(video, VE_CTRL, ~VE_CTRL_FRC, + FIELD_PREP(VE_CTRL_FRC, frame_rate)); + } + + return 0; +} + +static const struct v4l2_ioctl_ops aspeed_video_ioctls = { + .vidioc_querycap = aspeed_video_querycap, + .vidioc_g_fmt_vid_cap = aspeed_video_get_format, + .vidioc_s_fmt_vid_cap = aspeed_video_set_format, + .vidioc_g_jpegcomp = aspeed_video_get_jpegcomp, + .vidioc_s_jpegcomp = aspeed_video_set_jpegcomp, + .vidioc_g_parm = aspeed_video_get_parm, + .vidioc_s_parm = aspeed_video_set_parm, +}; + +static void aspeed_video_resolution_work(struct work_struct *work) +{ + int rc; + struct delayed_work *dwork = to_delayed_work(work); + struct aspeed_video *video = container_of(dwork, struct aspeed_video, + res_work); + + /* No clients remaining after delay */ + if (atomic_read(&video->clients) == 0) + goto done; + + aspeed_video_on(video); + + aspeed_video_init_regs(video); + + rc = aspeed_video_get_resolution(video); + if (rc) { + dev_err(video->dev, + "resolution changed; couldn't get new resolution\n"); + } else { + video->frame_idx = 0; + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); + } + +done: + clear_bit(VIDEO_RES_CHANGE, &video->flags); + wake_up_interruptible_all(&video->wait); +} + +static bool aspeed_video_frame_available(struct aspeed_video *video) +{ + if (!test_and_clear_bit(VIDEO_FRAME_AVAILABLE, &video->flags)) { + if (!test_bit(VIDEO_FRAME_TRIGGERED, &video->flags)) + aspeed_video_start_frame(video); + + return false; + } + + return true; +} + +static ssize_t aspeed_video_file_read(struct file *file, char __user *buf, + size_t count, loff_t *ppos) +{ + int rc; + int fidx; + size_t size; + struct aspeed_video *video = video_drvdata(file); + + if (mutex_lock_interruptible(&video->video_lock)) + return -EINTR; + + if (file->f_flags & O_NONBLOCK) { + if (!aspeed_video_frame_available(video)) { + rc = -EAGAIN; + goto unlock; + } else { + goto ready; + } + } + + rc = wait_event_interruptible(video->wait, + aspeed_video_frame_available(video)); + if (rc) { + rc = -EINTR; + goto unlock; + } + +ready: + fidx = video->frame_idx; + size = min_t(size_t, video->frame_size, count); + aspeed_video_start_frame(video); + + if (copy_to_user(buf, video->comp[fidx].virt, size)) { + rc = -EFAULT; + goto unlock; + } + + rc = size; + +unlock: + mutex_unlock(&video->video_lock); + return rc; +} + +static int aspeed_video_open(struct file *file) +{ + int rc; + struct aspeed_video *video = video_drvdata(file); + + if (atomic_inc_return(&video->clients) == 1) { + rc = aspeed_video_start(video); + if (rc) { + dev_err(video->dev, "Failed to start video engine\n"); + atomic_dec(&video->clients); + return rc; + } + } + + return v4l2_fh_open(file); +} + +static int aspeed_video_release(struct file *file) +{ + int rc; + struct aspeed_video *video = video_drvdata(file); + + rc = v4l2_fh_release(file); + + if (atomic_dec_return(&video->clients) == 0) + aspeed_video_stop(video); + + return rc; +} + +static const struct v4l2_file_operations aspeed_video_v4l2_fops = { + .owner = THIS_MODULE, + .read = aspeed_video_file_read, + .unlocked_ioctl = video_ioctl2, + .open = aspeed_video_open, + .release = aspeed_video_release, +}; + +static void aspeed_video_device_release(struct video_device *vdev) +{ +} + +static int aspeed_video_setup_video(struct aspeed_video *video) +{ + int rc; + struct v4l2_device *v4l2_dev = &video->v4l2_dev; + struct video_device *vdev = &video->vdev; + + rc = v4l2_device_register(video->dev, v4l2_dev); + if (rc) { + dev_err(video->dev, "Failed to register v4l2 device\n"); + return rc; + } + + vdev->fops = &aspeed_video_v4l2_fops; + vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_READWRITE; + vdev->v4l2_dev = v4l2_dev; + strncpy(vdev->name, DEVICE_NAME, sizeof(vdev->name)); + vdev->vfl_type = VFL_TYPE_GRABBER; + vdev->vfl_dir = VFL_DIR_RX; + vdev->release = aspeed_video_device_release; + vdev->ioctl_ops = &aspeed_video_ioctls; + vdev->lock = &video->video_lock; + + video_set_drvdata(vdev, video); + rc = video_register_device(vdev, VFL_TYPE_GRABBER, 0); + if (rc) { + v4l2_device_unregister(v4l2_dev); + dev_err(video->dev, "Failed to register video device\n"); + return rc; + } + + /* set pixel format defaults */ + video->fmt.pixelformat = V4L2_PIX_FMT_YUV444; + video->fmt.field = V4L2_FIELD_NONE; + + return 0; +} + +static int aspeed_video_init(struct aspeed_video *video) +{ + int irq; + int rc; + struct device *dev = video->dev; + + irq = irq_of_parse_and_map(dev->of_node, 0); + if (!irq) { + dev_err(dev, "Unable to find IRQ\n"); + return -ENODEV; + } + + rc = devm_request_irq(dev, irq, aspeed_video_irq, IRQF_SHARED, + DEVICE_NAME, video); + if (rc < 0) { + dev_err(dev, "Unable to request IRQ %d\n", irq); + return rc; + } + + video->eclk = devm_clk_get(dev, "eclk-gate"); + if (IS_ERR(video->eclk)) { + dev_err(dev, "Unable to get ECLK\n"); + return PTR_ERR(video->eclk); + } + + video->vclk = devm_clk_get(dev, "vclk-gate"); + if (IS_ERR(video->vclk)) { + dev_err(dev, "Unable to get VCLK\n"); + return PTR_ERR(video->vclk); + } + + video->rst = devm_reset_control_get_exclusive(dev, NULL); + if (IS_ERR(video->rst)) { + dev_err(dev, "Unable to get VE reset\n"); + return PTR_ERR(video->rst); + } + + rc = of_reserved_mem_device_init(dev); + if (rc) { + dev_err(dev, "Unable to reserve memory\n"); + return rc; + } + + rc = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32)); + if (rc) { + dev_err(dev, "Failed to set DMA mask\n"); + of_reserved_mem_device_release(dev); + return rc; + } + + return 0; +} + +static int aspeed_video_probe(struct platform_device *pdev) +{ + int rc; + struct resource *res; + struct aspeed_video *video = kzalloc(sizeof(*video), GFP_KERNEL); + + if (!video) + return -ENOMEM; + + video->frame_rate = 30; + video->dev = &pdev->dev; + mutex_init(&video->video_lock); + init_waitqueue_head(&video->wait); + INIT_DELAYED_WORK(&video->res_work, aspeed_video_resolution_work); + + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + + video->base = devm_ioremap_resource(video->dev, res); + + if (IS_ERR(video->base)) + return PTR_ERR(video->base); + + rc = aspeed_video_init(video); + if (rc) + return rc; + + rc = aspeed_video_setup_video(video); + if (rc) + return rc; + + return 0; +} + +static int aspeed_video_remove(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct v4l2_device *v4l2_dev = dev_get_drvdata(dev); + struct aspeed_video *video = to_aspeed_video(v4l2_dev); + + video_unregister_device(&video->vdev); + + v4l2_device_unregister(v4l2_dev); + + of_reserved_mem_device_release(dev); + + return 0; +} + +static const struct of_device_id aspeed_video_of_match[] = { + { .compatible = "aspeed,ast2400-video" }, + { .compatible = "aspeed,ast2500-video" }, + {} +}; +MODULE_DEVICE_TABLE(of, aspeed_video_of_match); + +static struct platform_driver aspeed_video_driver = { + .driver = { + .name = DEVICE_NAME, + .of_match_table = aspeed_video_of_match, + }, + .probe = aspeed_video_probe, + .remove = aspeed_video_remove, +}; + +module_platform_driver(aspeed_video_driver); + +MODULE_DESCRIPTION("ASPEED Video Engine Driver"); +MODULE_AUTHOR("Eddie James"); +MODULE_LICENSE("GPL v2"); -- 1.8.3.1 ^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 4/4] media: platform: Add Aspeed Video Engine driver 2018-08-29 21:09 ` [PATCH 4/4] media: platform: Add Aspeed Video Engine driver Eddie James @ 2018-08-30 0:52 ` Ezequiel Garcia 2018-08-30 15:40 ` Eddie James 2018-09-03 11:40 ` Hans Verkuil 1 sibling, 1 reply; 21+ messages in thread From: Ezequiel Garcia @ 2018-08-30 0:52 UTC (permalink / raw) To: linux-arm-kernel Hi Eddie, On 29 August 2018 at 18:09, Eddie James <eajames@linux.vnet.ibm.com> wrote: > The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs > can capture and compress video data from digital or analog sources. With > the Aspeed chip acting a service processor, the Video Engine can capture > the host processor graphics output. > > Add a V4L2 driver to capture video data and compress it to JPEG images, > making the data available through a standard read interface. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- > drivers/media/platform/Kconfig | 8 + > drivers/media/platform/Makefile | 1 + > drivers/media/platform/aspeed-video.c | 1307 +++++++++++++++++++++++++++++++++ > 3 files changed, 1316 insertions(+) > create mode 100644 drivers/media/platform/aspeed-video.c > > diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig > index 94c1fe0..e599245 100644 > --- a/drivers/media/platform/Kconfig > +++ b/drivers/media/platform/Kconfig > @@ -32,6 +32,14 @@ source "drivers/media/platform/davinci/Kconfig" > > source "drivers/media/platform/omap/Kconfig" > > +config VIDEO_ASPEED > + tristate "Aspeed AST2400 and AST2500 Video Engine driver" > + depends on VIDEO_V4L2 It seems you are not using videobuf2. I think it should simplify the read I/O part and at the same time expose the other capture methods. There are plenty of examples to follow. Regards, Eze ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 4/4] media: platform: Add Aspeed Video Engine driver 2018-08-30 0:52 ` Ezequiel Garcia @ 2018-08-30 15:40 ` Eddie James 0 siblings, 0 replies; 21+ messages in thread From: Eddie James @ 2018-08-30 15:40 UTC (permalink / raw) To: linux-arm-kernel On 08/29/2018 07:52 PM, Ezequiel Garcia wrote: > Hi Eddie, > > On 29 August 2018 at 18:09, Eddie James <eajames@linux.vnet.ibm.com> wrote: >> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >> can capture and compress video data from digital or analog sources. With >> the Aspeed chip acting a service processor, the Video Engine can capture >> the host processor graphics output. >> >> Add a V4L2 driver to capture video data and compress it to JPEG images, >> making the data available through a standard read interface. >> >> Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> >> --- >> drivers/media/platform/Kconfig | 8 + >> drivers/media/platform/Makefile | 1 + >> drivers/media/platform/aspeed-video.c | 1307 +++++++++++++++++++++++++++++++++ >> 3 files changed, 1316 insertions(+) >> create mode 100644 drivers/media/platform/aspeed-video.c >> >> diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig >> index 94c1fe0..e599245 100644 >> --- a/drivers/media/platform/Kconfig >> +++ b/drivers/media/platform/Kconfig >> @@ -32,6 +32,14 @@ source "drivers/media/platform/davinci/Kconfig" >> >> source "drivers/media/platform/omap/Kconfig" >> >> +config VIDEO_ASPEED >> + tristate "Aspeed AST2400 and AST2500 Video Engine driver" >> + depends on VIDEO_V4L2 > It seems you are not using videobuf2. I think it should simplify the read > I/O part and at the same time expose the other capture methods. Hi, Well I'm not sure it would simplify the read interface; it's quite simple as it is, both in the driver and to set up in user-space. I did get streaming I/O working but found the performance significantly worse than simple read calls, and therefore not worth the additional complexity. Is it required that I support streaming? Thanks, Eddie > > There are plenty of examples to follow. > > Regards, > Eze > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 4/4] media: platform: Add Aspeed Video Engine driver 2018-08-29 21:09 ` [PATCH 4/4] media: platform: Add Aspeed Video Engine driver Eddie James 2018-08-30 0:52 ` Ezequiel Garcia @ 2018-09-03 11:40 ` Hans Verkuil 2018-09-13 19:00 ` Eddie James 1 sibling, 1 reply; 21+ messages in thread From: Hans Verkuil @ 2018-09-03 11:40 UTC (permalink / raw) To: linux-arm-kernel On 08/29/2018 11:09 PM, Eddie James wrote: > The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs > can capture and compress video data from digital or analog sources. With > the Aspeed chip acting a service processor, the Video Engine can capture > the host processor graphics output. > > Add a V4L2 driver to capture video data and compress it to JPEG images, > making the data available through a standard read interface. > > Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> > --- > drivers/media/platform/Kconfig | 8 + > drivers/media/platform/Makefile | 1 + > drivers/media/platform/aspeed-video.c | 1307 +++++++++++++++++++++++++++++++++ Missing MAINTAINERS file update. > 3 files changed, 1316 insertions(+) > create mode 100644 drivers/media/platform/aspeed-video.c > > diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig > index 94c1fe0..e599245 100644 > --- a/drivers/media/platform/Kconfig > +++ b/drivers/media/platform/Kconfig > @@ -32,6 +32,14 @@ source "drivers/media/platform/davinci/Kconfig" > > source "drivers/media/platform/omap/Kconfig" > > +config VIDEO_ASPEED > + tristate "Aspeed AST2400 and AST2500 Video Engine driver" > + depends on VIDEO_V4L2 > + help > + Support for the Aspeed Video Engine (VE) embedded in the Aspeed > + AST2400 and AST2500 SOCs. The VE can capture and compress video data > + from digital or analog sources. > + > config VIDEO_SH_VOU > tristate "SuperH VOU video output driver" > depends on MEDIA_CAMERA_SUPPORT > diff --git a/drivers/media/platform/Makefile b/drivers/media/platform/Makefile > index 41322ab..205c33a 100644 > --- a/drivers/media/platform/Makefile > +++ b/drivers/media/platform/Makefile > @@ -3,6 +3,7 @@ > # Makefile for the video capture/playback device drivers. > # > > +obj-$(CONFIG_VIDEO_ASPEED) += aspeed-video.o > obj-$(CONFIG_VIDEO_CADENCE) += cadence/ > obj-$(CONFIG_VIDEO_VIA_CAMERA) += via-camera.o > obj-$(CONFIG_VIDEO_CAFE_CCIC) += marvell-ccic/ > diff --git a/drivers/media/platform/aspeed-video.c b/drivers/media/platform/aspeed-video.c > new file mode 100644 > index 0000000..4f36a3f > --- /dev/null > +++ b/drivers/media/platform/aspeed-video.c > @@ -0,0 +1,1307 @@ > +// SPDX-License-Identifier: GPL-2.0+ > +/* > + * ASPEED Video Engine Driver > + * > + * Copyright 2018 IBM Corp. > + * > + * Eddie James <eajames@linux.vnet.ibm.com> > + * > + * This program is free software; you can redistribute it and/or modify > + * it under the terms of the GNU General Public License version 2 as > + * published by the Free Software Foundation. > + */ > + > +#include <linux/atomic.h> > +#include <linux/bitfield.h> > +#include <linux/clk.h> > +#include <linux/delay.h> > +#include <linux/device.h> > +#include <linux/dma-mapping.h> > +#include <linux/interrupt.h> > +#include <linux/jiffies.h> > +#include <linux/module.h> > +#include <linux/mutex.h> > +#include <linux/of.h> > +#include <linux/of_irq.h> > +#include <linux/of_reserved_mem.h> > +#include <linux/platform_device.h> > +#include <linux/reset.h> > +#include <linux/sched.h> > +#include <linux/spinlock.h> > +#include <linux/string.h> > +#include <linux/videodev2.h> > +#include <linux/wait.h> > +#include <linux/workqueue.h> > +#include <media/v4l2-dev.h> > +#include <media/v4l2-device.h> > +#include <media/v4l2-ioctl.h> > + > +#define DEVICE_NAME "aspeed-video" > + > +#define ASPEED_VIDEO_JPEG_NUM_QUALITIES 12 > +#define ASPEED_VIDEO_JPEG_HEADER_SIZE 10 > +#define ASPEED_VIDEO_JPEG_QUANT_SIZE 116 > +#define ASPEED_VIDEO_JPEG_DCT_SIZE 34 > + > +#define NUM_POLARITY_CHECKS 10 > +#define INVALID_RESOLUTION_RETRIES 1 > +#define INVALID_RESOLUTION_DELAY msecs_to_jiffies(250) > +#define RESOLUTION_CHANGE_DELAY msecs_to_jiffies(500) > +#define MODE_DETECT_TIMEOUT msecs_to_jiffies(500) > +#define DIRECT_FETCH_THRESHOLD 0x0c0000 /* 1024 * 768, 32bpp */ > + > +#define VE_SRC_BUFFER_SIZE 0x900000 /* 1920 * 1200, 32bpp */ > +#define VE_COMP_BUFFER_SIZE 0x100000 /* 128K packet * 8 packets */ > +#define VE_JPEG_BUFFER_SIZE 0x006000 /* 512 * 12 * 4 */ > + > +#define VE_PROTECTION_KEY 0x000 > +#define VE_PROTECTION_KEY_UNLOCK 0x1A038AA8 > + > +#define VE_SEQ_CTRL 0x004 > +#define VE_SEQ_CTRL_TRIG_MODE_DET BIT(0) > +#define VE_SEQ_CTRL_TRIG_CAPTURE BIT(1) > +#define VE_SEQ_CTRL_FORCE_IDLE BIT(2) > +#define VE_SEQ_CTRL_MULT_FRAME BIT(3) > +#define VE_SEQ_CTRL_TRIG_COMP BIT(4) > +#define VE_SEQ_CTRL_AUTO_COMP BIT(5) > +#define VE_SEQ_CTRL_EN_WATCHDOG BIT(7) > +#define VE_SEQ_CTRL_YUV420 BIT(10) > +#define VE_SEQ_CTRL_COMP_FMT GENMASK(11, 10) > +#define VE_SEQ_CTRL_HALT BIT(12) > +#define VE_SEQ_CTRL_EN_WATCHDOG_COMP BIT(14) > +#define VE_SEQ_CTRL_TRIG_JPG BIT(15) > +#define VE_SEQ_CTRL_CAP_BUSY BIT(16) > +#define VE_SEQ_CTRL_COMP_BUSY BIT(18) > + > +#ifdef CONFIG_MACH_ASPEED_G5 > +#define VE_SEQ_CTRL_JPEG_MODE BIT(13) /* AST2500 */ > +#else > +#define VE_SEQ_CTRL_JPEG_MODE BIT(8) /* AST2400 */ > +#endif /* CONFIG_MACH_ASPEED_G5 */ > + > +#define VE_CTRL 0x008 > +#define VE_CTRL_HSYNC_POL BIT(0) > +#define VE_CTRL_VSYNC_POL BIT(1) > +#define VE_CTRL_SOURCE BIT(2) > +#define VE_CTRL_INT_DE BIT(4) > +#define VE_CTRL_DIRECT_FETCH BIT(5) > +#define VE_CTRL_YUV BIT(6) > +#define VE_CTRL_RGB BIT(7) > +#define VE_CTRL_CAPTURE_FMT GENMASK(7, 6) > +#define VE_CTRL_AUTO_OR_CURSOR BIT(8) > +#define VE_CTRL_CLK_INVERSE BIT(11) > +#define VE_CTRL_CLK_DELAY GENMASK(11, 9) > +#define VE_CTRL_INTERLACE BIT(14) > +#define VE_CTRL_HSYNC_POL_CTRL BIT(15) > +#define VE_CTRL_FRC GENMASK(23, 16) > + > +#define VE_TGS_0 0x00c > +#define VE_TGS_1 0x010 > +#define VE_TGS_FIRST GENMASK(28, 16) > +#define VE_TGS_LAST GENMASK(12, 0) > + > +#define VE_SCALING_FACTOR 0x014 > +#define VE_SCALING_FILTER0 0x018 > +#define VE_SCALING_FILTER1 0x01c > +#define VE_SCALING_FILTER2 0x020 > +#define VE_SCALING_FILTER3 0x024 > + > +#define VE_CAP_WINDOW 0x030 > +#define VE_COMP_WINDOW 0x034 > +#define VE_COMP_PROC_OFFSET 0x038 > +#define VE_COMP_OFFSET 0x03c > +#define VE_JPEG_ADDR 0x040 > +#define VE_SRC0_ADDR 0x044 > +#define VE_SRC_SCANLINE_OFFSET 0x048 > +#define VE_SRC1_ADDR 0x04c > +#define VE_COMP_ADDR 0x054 > + > +#define VE_STREAM_BUF_SIZE 0x058 > +#define VE_STREAM_BUF_SIZE_N_PACKETS GENMASK(5, 3) > +#define VE_STREAM_BUF_SIZE_P_SIZE GENMASK(2, 0) > + > +#define VE_COMP_CTRL 0x060 > +#define VE_COMP_CTRL_VQ_DCT_ONLY BIT(0) > +#define VE_COMP_CTRL_VQ_4COLOR BIT(1) > +#define VE_COMP_CTRL_QUANTIZE BIT(2) > +#define VE_COMP_CTRL_EN_BQ BIT(4) > +#define VE_COMP_CTRL_EN_CRYPTO BIT(5) > +#define VE_COMP_CTRL_DCT_CHR GENMASK(10, 6) > +#define VE_COMP_CTRL_DCT_LUM GENMASK(15, 11) > +#define VE_COMP_CTRL_EN_HQ BIT(16) > +#define VE_COMP_CTRL_RSVD BIT(19) > +#define VE_COMP_CTRL_ENCODE GENMASK(21, 20) > +#define VE_COMP_CTRL_HQ_DCT_CHR GENMASK(26, 22) > +#define VE_COMP_CTRL_HQ_DCT_LUM GENMASK(31, 27) > + > +#define VE_OFFSET_COMP_STREAM 0x078 > + > +#define VE_SRC_LR_EDGE_DET 0x090 > +#define VE_SRC_LR_EDGE_DET_LEFT GENMASK(11, 0) > +#define VE_SRC_LR_EDGE_DET_NO_V BIT(12) > +#define VE_SRC_LR_EDGE_DET_NO_H BIT(13) > +#define VE_SRC_LR_EDGE_DET_NO_DISP BIT(14) > +#define VE_SRC_LR_EDGE_DET_NO_CLK BIT(15) > +#define VE_SRC_LR_EDGE_DET_RT_SHF 16 > +#define VE_SRC_LR_EDGE_DET_RT GENMASK(27, VE_SRC_LR_EDGE_DET_RT_SHF) > +#define VE_SRC_LR_EDGE_DET_INTERLACE BIT(31) > + > +#define VE_SRC_TB_EDGE_DET 0x094 > +#define VE_SRC_TB_EDGE_DET_TOP GENMASK(12, 0) > +#define VE_SRC_TB_EDGE_DET_BOT_SHF 16 > +#define VE_SRC_TB_EDGE_DET_BOT GENMASK(28, VE_SRC_TB_EDGE_DET_BOT_SHF) > + > +#define VE_MODE_DETECT_STATUS 0x098 > +#define VE_MODE_DETECT_STATUS_VSYNC BIT(28) > +#define VE_MODE_DETECT_STATUS_HSYNC BIT(29) > + > +#define VE_INTERRUPT_CTRL 0x304 > +#define VE_INTERRUPT_STATUS 0x308 > +#define VE_INTERRUPT_MODE_DETECT_WD BIT(0) > +#define VE_INTERRUPT_CAPTURE_COMPLETE BIT(1) > +#define VE_INTERRUPT_COMP_READY BIT(2) > +#define VE_INTERRUPT_COMP_COMPLETE BIT(3) > +#define VE_INTERRUPT_MODE_DETECT BIT(4) > +#define VE_INTERRUPT_FRAME_COMPLETE BIT(5) > +#define VE_INTERRUPT_DECODE_ERR BIT(6) > +#define VE_INTERRUPT_HALT_READY BIT(8) > +#define VE_INTERRUPT_HANG_WD BIT(9) > +#define VE_INTERRUPT_STREAM_DESC BIT(10) > +#define VE_INTERRUPT_VSYNC_DESC BIT(11) > + > +#define VE_MODE_DETECT 0x30c > +#define VE_MEM_RESTRICT_START 0x310 > +#define VE_MEM_RESTRICT_END 0x314 > + > +enum { > + VIDEO_MODE_DETECT_DONE, > + VIDEO_RES_CHANGE, > + VIDEO_FRAME_AVAILABLE, > + VIDEO_FRAME_TRIGGERED, > +}; > + > +struct aspeed_video_addr { > + dma_addr_t dma; > + void *virt; > +}; > + > +struct aspeed_video { > + void __iomem *base; > + struct clk *eclk; > + struct clk *vclk; > + struct reset_control *rst; > + > + struct device *dev; > + struct v4l2_device v4l2_dev; > + struct video_device vdev; > + struct mutex video_lock; > + > + atomic_t clients; > + wait_queue_head_t wait; > + struct delayed_work res_work; > + unsigned long flags; > + > + int frame_idx; > + u32 frame_size; > + > + dma_addr_t max; > + dma_addr_t min; > + struct aspeed_video_addr srcs[2]; > + struct aspeed_video_addr comp[2]; > + struct aspeed_video_addr jpeg; > + > + int frame_rate; > + int jpeg_quality; > + struct v4l2_pix_format fmt; > +}; > + > +#define to_aspeed_video(x) container_of((x), struct aspeed_video, v4l2_dev) > + > +static const u32 aspeed_video_jpeg_header[ASPEED_VIDEO_JPEG_HEADER_SIZE] = { > + 0xE0FFD8FF, 0x464A1000, 0x01004649, 0x60000101, 0x00006000, 0x0F00FEFF, > + 0x00002D05, 0x00000000, 0x00000000, 0x00DBFF00 > +}; > + > +static const u32 aspeed_video_jpeg_quant[ASPEED_VIDEO_JPEG_QUANT_SIZE] = { > + 0x081100C0, 0x00000000, 0x00110103, 0x03011102, 0xC4FF0111, 0x00001F00, > + 0x01010501, 0x01010101, 0x00000000, 0x00000000, 0x04030201, 0x08070605, > + 0xFF0B0A09, 0x10B500C4, 0x03010200, 0x03040203, 0x04040505, 0x7D010000, > + 0x00030201, 0x12051104, 0x06413121, 0x07615113, 0x32147122, 0x08A19181, > + 0xC1B14223, 0xF0D15215, 0x72623324, 0x160A0982, 0x1A191817, 0x28272625, > + 0x35342A29, 0x39383736, 0x4544433A, 0x49484746, 0x5554534A, 0x59585756, > + 0x6564635A, 0x69686766, 0x7574736A, 0x79787776, 0x8584837A, 0x89888786, > + 0x9493928A, 0x98979695, 0xA3A29A99, 0xA7A6A5A4, 0xB2AAA9A8, 0xB6B5B4B3, > + 0xBAB9B8B7, 0xC5C4C3C2, 0xC9C8C7C6, 0xD4D3D2CA, 0xD8D7D6D5, 0xE2E1DAD9, > + 0xE6E5E4E3, 0xEAE9E8E7, 0xF4F3F2F1, 0xF8F7F6F5, 0xC4FFFAF9, 0x00011F00, > + 0x01010103, 0x01010101, 0x00000101, 0x00000000, 0x04030201, 0x08070605, > + 0xFF0B0A09, 0x11B500C4, 0x02010200, 0x04030404, 0x04040507, 0x77020100, > + 0x03020100, 0x21050411, 0x41120631, 0x71610751, 0x81322213, 0x91421408, > + 0x09C1B1A1, 0xF0523323, 0xD1726215, 0x3424160A, 0x17F125E1, 0x261A1918, > + 0x2A292827, 0x38373635, 0x44433A39, 0x48474645, 0x54534A49, 0x58575655, > + 0x64635A59, 0x68676665, 0x74736A69, 0x78777675, 0x83827A79, 0x87868584, > + 0x928A8988, 0x96959493, 0x9A999897, 0xA5A4A3A2, 0xA9A8A7A6, 0xB4B3B2AA, > + 0xB8B7B6B5, 0xC3C2BAB9, 0xC7C6C5C4, 0xD2CAC9C8, 0xD6D5D4D3, 0xDAD9D8D7, > + 0xE5E4E3E2, 0xE9E8E7E6, 0xF4F3F2EA, 0xF8F7F6F5, 0xDAFFFAF9, 0x01030C00, > + 0x03110200, 0x003F0011 > +}; > + > +static const u32 aspeed_video_jpeg_dct[ASPEED_VIDEO_JPEG_NUM_QUALITIES] > + [ASPEED_VIDEO_JPEG_DCT_SIZE] = { > + { 0x0D140043, 0x0C0F110F, 0x11101114, 0x17141516, 0x1E20321E, > + 0x3D1E1B1B, 0x32242E2B, 0x4B4C3F48, 0x44463F47, 0x61735A50, > + 0x566C5550, 0x88644644, 0x7A766C65, 0x4D808280, 0x8C978D60, > + 0x7E73967D, 0xDBFF7B80, 0x1F014300, 0x272D2121, 0x3030582D, > + 0x697BB958, 0xB8B9B97B, 0xB9B8A6A6, 0xB9B9B9B9, 0xB9B9B9B9, > + 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, > + 0xB9B9B9B9, 0xB9B9B9B9, 0xB9B9B9B9, 0xFFB9B9B9 }, > + { 0x0C110043, 0x0A0D0F0D, 0x0F0E0F11, 0x14111213, 0x1A1C2B1A, > + 0x351A1818, 0x2B1F2826, 0x4142373F, 0x3C3D373E, 0x55644E46, > + 0x4B5F4A46, 0x77573D3C, 0x6B675F58, 0x43707170, 0x7A847B54, > + 0x6E64836D, 0xDBFF6C70, 0x1B014300, 0x22271D1D, 0x2A2A4C27, > + 0x5B6BA04C, 0xA0A0A06B, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, > + 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, > + 0xA0A0A0A0, 0xA0A0A0A0, 0xA0A0A0A0, 0xFFA0A0A0 }, > + { 0x090E0043, 0x090A0C0A, 0x0C0B0C0E, 0x110E0F10, 0x15172415, > + 0x2C151313, 0x241A211F, 0x36372E34, 0x31322E33, 0x4653413A, > + 0x3E4E3D3A, 0x62483231, 0x58564E49, 0x385D5E5D, 0x656D6645, > + 0x5B536C5A, 0xDBFF595D, 0x16014300, 0x1C201818, 0x22223F20, > + 0x4B58853F, 0x85858558, 0x85858585, 0x85858585, 0x85858585, > + 0x85858585, 0x85858585, 0x85858585, 0x85858585, 0x85858585, > + 0x85858585, 0x85858585, 0x85858585, 0xFF858585 }, > + { 0x070B0043, 0x07080A08, 0x0A090A0B, 0x0D0B0C0C, 0x11121C11, > + 0x23110F0F, 0x1C141A19, 0x2B2B2429, 0x27282428, 0x3842332E, > + 0x313E302E, 0x4E392827, 0x46443E3A, 0x2C4A4A4A, 0x50565137, > + 0x48425647, 0xDBFF474A, 0x12014300, 0x161A1313, 0x1C1C331A, > + 0x3D486C33, 0x6C6C6C48, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, > + 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, > + 0x6C6C6C6C, 0x6C6C6C6C, 0x6C6C6C6C, 0xFF6C6C6C }, > + { 0x06090043, 0x05060706, 0x07070709, 0x0A09090A, 0x0D0E160D, > + 0x1B0D0C0C, 0x16101413, 0x21221C20, 0x1E1F1C20, 0x2B332824, > + 0x26302624, 0x3D2D1F1E, 0x3735302D, 0x22393A39, 0x3F443F2B, > + 0x38334338, 0xDBFF3739, 0x0D014300, 0x11130E0E, 0x15152613, > + 0x2D355026, 0x50505035, 0x50505050, 0x50505050, 0x50505050, > + 0x50505050, 0x50505050, 0x50505050, 0x50505050, 0x50505050, > + 0x50505050, 0x50505050, 0x50505050, 0xFF505050 }, > + { 0x04060043, 0x03040504, 0x05040506, 0x07060606, 0x09090F09, > + 0x12090808, 0x0F0A0D0D, 0x16161315, 0x14151315, 0x1D221B18, > + 0x19201918, 0x281E1514, 0x2423201E, 0x17262726, 0x2A2D2A1C, > + 0x25222D25, 0xDBFF2526, 0x09014300, 0x0B0D0A0A, 0x0E0E1A0D, > + 0x1F25371A, 0x37373725, 0x37373737, 0x37373737, 0x37373737, > + 0x37373737, 0x37373737, 0x37373737, 0x37373737, 0x37373737, > + 0x37373737, 0x37373737, 0x37373737, 0xFF373737 }, > + { 0x02030043, 0x01020202, 0x02020203, 0x03030303, 0x04040704, > + 0x09040404, 0x07050606, 0x0B0B090A, 0x0A0A090A, 0x0E110D0C, > + 0x0C100C0C, 0x140F0A0A, 0x1211100F, 0x0B131313, 0x1516150E, > + 0x12111612, 0xDBFF1213, 0x04014300, 0x05060505, 0x07070D06, > + 0x0F121B0D, 0x1B1B1B12, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, > + 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, > + 0x1B1B1B1B, 0x1B1B1B1B, 0x1B1B1B1B, 0xFF1B1B1B }, > + { 0x01020043, 0x01010101, 0x01010102, 0x02020202, 0x03030503, > + 0x06030202, 0x05030404, 0x07070607, 0x06070607, 0x090B0908, > + 0x080A0808, 0x0D0A0706, 0x0C0B0A0A, 0x070C0D0C, 0x0E0F0E09, > + 0x0C0B0F0C, 0xDBFF0C0C, 0x03014300, 0x03040303, 0x04040804, > + 0x0A0C1208, 0x1212120C, 0x12121212, 0x12121212, 0x12121212, > + 0x12121212, 0x12121212, 0x12121212, 0x12121212, 0x12121212, > + 0x12121212, 0x12121212, 0x12121212, 0xFF121212 }, > + { 0x01020043, 0x01010101, 0x01010102, 0x02020202, 0x03030503, > + 0x06030202, 0x05030404, 0x07070607, 0x06070607, 0x090B0908, > + 0x080A0808, 0x0D0A0706, 0x0C0B0A0A, 0x070C0D0C, 0x0E0F0E09, > + 0x0C0B0F0C, 0xDBFF0C0C, 0x02014300, 0x03030202, 0x04040703, > + 0x080A0F07, 0x0F0F0F0A, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, > + 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, > + 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0xFF0F0F0F }, > + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x02020302, > + 0x04020202, 0x03020303, 0x05050405, 0x05050405, 0x07080606, > + 0x06080606, 0x0A070505, 0x09080807, 0x05090909, 0x0A0B0A07, > + 0x09080B09, 0xDBFF0909, 0x02014300, 0x02030202, 0x03030503, > + 0x07080C05, 0x0C0C0C08, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, > + 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, > + 0x0C0C0C0C, 0x0C0C0C0C, 0x0C0C0C0C, 0xFF0C0C0C }, > + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x01010201, > + 0x03010101, 0x02010202, 0x03030303, 0x03030303, 0x04050404, > + 0x04050404, 0x06050303, 0x06050505, 0x03060606, 0x07070704, > + 0x06050706, 0xDBFF0606, 0x01014300, 0x01020101, 0x02020402, > + 0x05060904, 0x09090906, 0x09090909, 0x09090909, 0x09090909, > + 0x09090909, 0x09090909, 0x09090909, 0x09090909, 0x09090909, > + 0x09090909, 0x09090909, 0x09090909, 0xFF090909 }, > + { 0x01010043, 0x01010101, 0x01010101, 0x01010101, 0x01010101, > + 0x01010101, 0x01010101, 0x01010101, 0x01010101, 0x02020202, > + 0x02020202, 0x03020101, 0x03020202, 0x01030303, 0x03030302, > + 0x03020303, 0xDBFF0403, 0x01014300, 0x01010101, 0x01010201, > + 0x03040602, 0x06060604, 0x06060606, 0x06060606, 0x06060606, > + 0x06060606, 0x06060606, 0x06060606, 0x06060606, 0x06060606, > + 0x06060606, 0x06060606, 0x06060606, 0xFF060606 } > +}; > + > +static void aspeed_video_init_jpeg_table(u32 *table, bool yuv420) > +{ > + int i; > + unsigned int base; > + > + for (i = 0; i < ASPEED_VIDEO_JPEG_NUM_QUALITIES; i++) { > + int j; > + > + base = 256 * i; /* AST HW requires this header spacing */ > + > + for (j = 0; j < ASPEED_VIDEO_JPEG_HEADER_SIZE; j++) > + table[base + j] = > + le32_to_cpu(aspeed_video_jpeg_header[j]); This doesn't look right. These tables are in cpu format to begin with, so le32_to_cpu doesn't make sense. BTW, what is the endianness of an aspeed SoC? > + > + base += ASPEED_VIDEO_JPEG_HEADER_SIZE; > + for (j = 0; j < ASPEED_VIDEO_JPEG_DCT_SIZE; j++) > + table[base + j] = > + le32_to_cpu(aspeed_video_jpeg_dct[i][j]); > + > + base += ASPEED_VIDEO_JPEG_DCT_SIZE; > + for (j = 0; j < ASPEED_VIDEO_JPEG_QUANT_SIZE; j++) > + table[base + j] = > + le32_to_cpu(aspeed_video_jpeg_quant[j]); > + > + if (yuv420) > + table[base + 2] = le32_to_cpu(0x00220103); > + } > +} > + > +static void aspeed_video_update(struct aspeed_video *video, u32 reg, > + unsigned long mask, u32 bits) > +{ > + u32 t = readl(video->base + reg); > + u32 before = t; > + > + t &= mask; > + t |= bits; > + writel(t, video->base + reg); > + dev_dbg(video->dev, "update %03x[%08x -> %08x]\n", reg, before, > + readl(video->base + reg)); > +} > + > +static u32 aspeed_video_read(struct aspeed_video *video, u32 reg) > +{ > + u32 t = readl(video->base + reg); > + > + dev_dbg(video->dev, "read %03x[%08x]\n", reg, t); > + return t; > +} > + > +static void aspeed_video_write(struct aspeed_video *video, u32 reg, u32 val) > +{ > + writel(val, video->base + reg); > + dev_dbg(video->dev, "write %03x[%08x]\n", reg, > + readl(video->base + reg)); > +} > + > +static bool aspeed_video_engine_busy(struct aspeed_video *video) > +{ > + u32 seq_ctrl = aspeed_video_read(video, VE_SEQ_CTRL); > + > + if (!(seq_ctrl & VE_SEQ_CTRL_COMP_BUSY) || > + !(seq_ctrl & VE_SEQ_CTRL_CAP_BUSY)) { > + dev_info(video->dev, "video engine busy\n"); > + return true; > + } > + > + return false; > +} > + > +static int aspeed_video_start_frame(struct aspeed_video *video) > +{ > + if (aspeed_video_engine_busy(video)) > + return -EBUSY; > + > + video->frame_idx = (video->frame_idx + 1) % 2; > + > + aspeed_video_write(video, VE_COMP_PROC_OFFSET, 0); > + aspeed_video_write(video, VE_COMP_OFFSET, 0); > + aspeed_video_write(video, VE_COMP_ADDR, > + video->comp[video->frame_idx].dma); > + > + set_bit(VIDEO_FRAME_TRIGGERED, &video->flags); > + > + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, > + VE_INTERRUPT_COMP_COMPLETE | > + VE_INTERRUPT_CAPTURE_COMPLETE); > + > + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, > + VE_SEQ_CTRL_TRIG_CAPTURE | VE_SEQ_CTRL_TRIG_COMP); > + > + return 0; > +} > + > +static void aspeed_video_start_mode_detect(struct aspeed_video *video) > +{ > + /* Enable mode detect interrupts */ > + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, > + VE_INTERRUPT_MODE_DETECT); > + > + /* Trigger mode detect */ > + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, > + VE_SEQ_CTRL_TRIG_MODE_DET); > +} > + > +static void aspeed_video_disable_mode_detect(struct aspeed_video *video) > +{ > + /* Disable mode detect interrupts */ > + aspeed_video_update(video, VE_INTERRUPT_CTRL, > + ~VE_INTERRUPT_MODE_DETECT, 0); > + > + /* Disable mode detect */ > + aspeed_video_update(video, VE_SEQ_CTRL, ~VE_SEQ_CTRL_TRIG_MODE_DET, 0); > +} > + > +static void aspeed_video_off(struct aspeed_video *video) > +{ > + /* Reset the engine */ > + reset_control_assert(video->rst); > + udelay(100); > + reset_control_deassert(video->rst); > + > + /* Turn off the relevant clocks */ > + clk_disable_unprepare(video->vclk); > + clk_disable_unprepare(video->eclk); > +} > + > +static void aspeed_video_on(struct aspeed_video *video) > +{ > + /* Turn on the relevant clocks */ > + clk_prepare_enable(video->eclk); > + clk_prepare_enable(video->vclk); > + > + /* Reset the engine */ > + reset_control_assert(video->rst); > + udelay(100); > + reset_control_deassert(video->rst); > +} > + > +static irqreturn_t aspeed_video_irq(int irq, void *arg) > +{ > + struct aspeed_video *video = arg; > + u32 sts = aspeed_video_read(video, VE_INTERRUPT_STATUS); > + > + if (atomic_read(&video->clients) == 0) { > + dev_info(video->dev, "irq with no client; disabling irqs\n"); > + > + aspeed_video_write(video, VE_INTERRUPT_CTRL, 0); > + aspeed_video_write(video, VE_INTERRUPT_STATUS, 0xFFFFFFFF); > + return IRQ_HANDLED; > + } > + > + /* Resolution changed; reset entire engine and reinitialize */ > + if (sts & VE_INTERRUPT_MODE_DETECT_WD) { > + dev_info(video->dev, "resolution changed; resetting\n"); > + set_bit(VIDEO_RES_CHANGE, &video->flags); > + > + aspeed_video_off(video); > + > + schedule_delayed_work(&video->res_work, > + RESOLUTION_CHANGE_DELAY); > + return IRQ_HANDLED; > + } > + > + if (sts & VE_INTERRUPT_MODE_DETECT) { > + aspeed_video_update(video, VE_INTERRUPT_CTRL, > + ~VE_INTERRUPT_MODE_DETECT, 0); > + aspeed_video_write(video, VE_INTERRUPT_STATUS, > + VE_INTERRUPT_MODE_DETECT); > + > + set_bit(VIDEO_MODE_DETECT_DONE, &video->flags); > + wake_up_interruptible_all(&video->wait); > + } > + > + if ((sts & VE_INTERRUPT_COMP_COMPLETE) && > + (sts & VE_INTERRUPT_CAPTURE_COMPLETE)) { > + video->frame_size = aspeed_video_read(video, > + VE_OFFSET_COMP_STREAM); > + > + aspeed_video_update(video, VE_INTERRUPT_CTRL, > + ~(VE_INTERRUPT_COMP_COMPLETE | > + VE_INTERRUPT_CAPTURE_COMPLETE), 0); > + aspeed_video_write(video, VE_INTERRUPT_STATUS, > + VE_INTERRUPT_COMP_COMPLETE | > + VE_INTERRUPT_CAPTURE_COMPLETE); > + aspeed_video_update(video, VE_SEQ_CTRL, > + ~(VE_SEQ_CTRL_TRIG_CAPTURE | > + VE_SEQ_CTRL_FORCE_IDLE | > + VE_SEQ_CTRL_TRIG_COMP), 0); > + > + set_bit(VIDEO_FRAME_AVAILABLE, &video->flags); > + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); > + wake_up_interruptible_all(&video->wait); > + } > + > + return IRQ_HANDLED; > +} > + > +static void aspeed_video_check_polarity(struct aspeed_video *video) > +{ > + int i; > + int hsync_counter = 0; > + int vsync_counter = 0; > + u32 sts; > + > + for (i = 0; i < NUM_POLARITY_CHECKS; ++i) { > + sts = aspeed_video_read(video, VE_MODE_DETECT_STATUS); > + if (sts & VE_MODE_DETECT_STATUS_VSYNC) > + vsync_counter--; > + else > + vsync_counter++; > + > + if (sts & VE_MODE_DETECT_STATUS_HSYNC) > + hsync_counter--; > + else > + hsync_counter++; > + } > + > + if (hsync_counter < 0 || vsync_counter < 0) { > + u32 ctrl; > + > + if (hsync_counter < 0) > + ctrl = VE_CTRL_HSYNC_POL; > + > + if (vsync_counter < 0) > + ctrl = VE_CTRL_VSYNC_POL; > + > + aspeed_video_update(video, VE_CTRL, 0xFFFFFFFF, ctrl); > + } > +} > + > +#define res_check(v) test_and_clear_bit(VIDEO_MODE_DETECT_DONE, &(v)->flags) > + > +static int aspeed_video_get_resolution(struct aspeed_video *video) > +{ > + bool invalid_resolution = true; > + int rc; > + int tries = 0; > + unsigned int bottom; > + unsigned int left; > + unsigned int right; > + unsigned int top; > + u32 src_lr_edge; > + u32 src_tb_edge; > + > + video->fmt.width = 0; > + video->fmt.height = 0; > + > + do { > + if (tries) { > + set_current_state(TASK_INTERRUPTIBLE); > + if (schedule_timeout(INVALID_RESOLUTION_DELAY)) > + return -EINTR; > + } > + > + aspeed_video_start_mode_detect(video); > + > + rc = wait_event_interruptible_timeout(video->wait, > + res_check(video), > + MODE_DETECT_TIMEOUT); > + if (!rc) { > + dev_err(video->dev, "timed out on 1st mode detect\n"); > + aspeed_video_disable_mode_detect(video); > + return -ETIME; > + } > + > + /* Disable mode detect in order to re-trigger */ > + aspeed_video_update(video, VE_SEQ_CTRL, > + ~VE_SEQ_CTRL_TRIG_MODE_DET, 0); > + > + aspeed_video_check_polarity(video); > + > + aspeed_video_start_mode_detect(video); > + > + rc = wait_event_interruptible_timeout(video->wait, > + res_check(video), > + MODE_DETECT_TIMEOUT); > + if (!rc) { > + dev_err(video->dev, "timed out on 2nd mode detect\n"); > + aspeed_video_disable_mode_detect(video); > + return -ETIME; > + } > + > + src_lr_edge = aspeed_video_read(video, VE_SRC_LR_EDGE_DET); > + src_tb_edge = aspeed_video_read(video, VE_SRC_TB_EDGE_DET); > + > + bottom = (src_tb_edge & VE_SRC_TB_EDGE_DET_BOT) >> > + VE_SRC_TB_EDGE_DET_BOT_SHF; > + top = src_tb_edge & VE_SRC_TB_EDGE_DET_TOP; > + if (top > bottom) > + continue; > + > + right = (src_lr_edge & VE_SRC_LR_EDGE_DET_RT) >> > + VE_SRC_LR_EDGE_DET_RT_SHF; > + left = src_lr_edge & VE_SRC_LR_EDGE_DET_LEFT; > + if (left > right) > + continue; > + > + invalid_resolution = false; > + } while (invalid_resolution && (tries++ < INVALID_RESOLUTION_RETRIES)); > + > + if (invalid_resolution) { > + dev_err(video->dev, "invalid resolution detected\n"); > + return -EMSGSIZE; > + } > + > + video->fmt.height = (bottom - top) + 1; > + video->fmt.width = (right - left) + 1; > + > + /* Don't use direct mode below 1024 x 768 (irqs don't fire) */ > + if (video->fmt.height * video->fmt.width < DIRECT_FETCH_THRESHOLD) { > + aspeed_video_write(video, VE_TGS_0, > + FIELD_PREP(VE_TGS_FIRST, left - 1) | > + FIELD_PREP(VE_TGS_LAST, right)); > + aspeed_video_write(video, VE_TGS_1, > + FIELD_PREP(VE_TGS_FIRST, top) | > + FIELD_PREP(VE_TGS_LAST, bottom + 1)); > + aspeed_video_update(video, VE_CTRL, > + ~VE_CTRL_DIRECT_FETCH, VE_CTRL_INT_DE); > + } > + > + aspeed_video_write(video, VE_CAP_WINDOW, > + video->fmt.width << 16 | video->fmt.height); > + aspeed_video_write(video, VE_COMP_WINDOW, > + video->fmt.width << 16 | video->fmt.height); > + aspeed_video_write(video, VE_SRC_SCANLINE_OFFSET, > + video->fmt.width * 4); > + > + aspeed_video_update(video, VE_INTERRUPT_CTRL, 0xFFFFFFFF, > + VE_INTERRUPT_MODE_DETECT_WD); > + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, > + VE_SEQ_CTRL_EN_WATCHDOG); > + > + dev_dbg(video->dev, "got resolution[%dx%d]\n", video->fmt.width, > + video->fmt.height); > + > + return 0; > +} > + > +static void aspeed_video_init_regs(struct aspeed_video *video) > +{ > + u32 comp_ctrl = VE_COMP_CTRL_RSVD | > + FIELD_PREP(VE_COMP_CTRL_DCT_LUM, video->jpeg_quality) | > + FIELD_PREP(VE_COMP_CTRL_DCT_CHR, video->jpeg_quality | 0x10); > + u32 ctrl = VE_CTRL_DIRECT_FETCH | VE_CTRL_AUTO_OR_CURSOR; > + u32 seq_ctrl = VE_SEQ_CTRL_AUTO_COMP | VE_SEQ_CTRL_JPEG_MODE; > + > + if (video->frame_rate) > + ctrl |= FIELD_PREP(VE_CTRL_FRC, video->frame_rate); > + > + if (video->fmt.pixelformat == V4L2_PIX_FMT_YUV420) > + seq_ctrl |= VE_SEQ_CTRL_YUV420; > + > + /* Unlock VE registers */ > + aspeed_video_write(video, VE_PROTECTION_KEY, VE_PROTECTION_KEY_UNLOCK); > + > + /* Disable interrupts */ > + aspeed_video_write(video, VE_INTERRUPT_CTRL, 0); > + aspeed_video_write(video, VE_INTERRUPT_STATUS, 0xFFFFFFFF); > + > + /* Clear the offset */ > + aspeed_video_write(video, VE_COMP_PROC_OFFSET, 0); > + aspeed_video_write(video, VE_COMP_OFFSET, 0); > + > + /* Set memory restrictions */ > + aspeed_video_write(video, VE_MEM_RESTRICT_START, video->min); > + aspeed_video_write(video, VE_MEM_RESTRICT_END, video->max); > + > + /* Set buffer addresses */ > + aspeed_video_write(video, VE_SRC0_ADDR, video->srcs[0].dma); > + aspeed_video_write(video, VE_SRC1_ADDR, video->srcs[1].dma); > + aspeed_video_write(video, VE_COMP_ADDR, video->comp[0].dma); > + aspeed_video_write(video, VE_JPEG_ADDR, video->jpeg.dma); > + > + /* Set control registers */ > + aspeed_video_write(video, VE_SEQ_CTRL, seq_ctrl); > + aspeed_video_write(video, VE_CTRL, ctrl); > + aspeed_video_write(video, VE_COMP_CTRL, comp_ctrl); > + > + /* Compression buffer size; 128K packet * 8 packets */ > + aspeed_video_write(video, VE_STREAM_BUF_SIZE, 0xf); > + > + /* Don't downscale */ > + aspeed_video_write(video, VE_SCALING_FACTOR, 0x10001000); > + aspeed_video_write(video, VE_SCALING_FILTER0, 0x00200000); > + aspeed_video_write(video, VE_SCALING_FILTER1, 0x00200000); > + aspeed_video_write(video, VE_SCALING_FILTER2, 0x00200000); > + aspeed_video_write(video, VE_SCALING_FILTER3, 0x00200000); > + > + /* Set mode detection defaults */ > + aspeed_video_write(video, VE_MODE_DETECT, 0x22666500); > +} > + > +static int aspeed_video_allocate_cma(struct aspeed_video *video) > +{ > + video->srcs[0].virt = dma_alloc_coherent(video->dev, > + VE_SRC_BUFFER_SIZE, > + &video->srcs[0].dma, > + GFP_KERNEL); > + if (!video->srcs[0].virt) { > + dev_err(video->dev, > + "Failed to allocate source buffer 0, size[%x]\n", > + VE_SRC_BUFFER_SIZE); > + goto err; > + } > + > + video->srcs[1].virt = dma_alloc_coherent(video->dev, > + VE_SRC_BUFFER_SIZE, > + &video->srcs[1].dma, > + GFP_KERNEL); > + if (!video->srcs[1].virt) { > + dev_err(video->dev, > + "Failed to allocate source buffer 1, size[%x]\n", > + VE_SRC_BUFFER_SIZE); > + goto free_src0; > + } > + > + video->comp[0].virt = dma_alloc_coherent(video->dev, > + VE_COMP_BUFFER_SIZE, > + &video->comp[0].dma, > + GFP_KERNEL); > + if (!video->comp[0].virt) { > + dev_err(video->dev, > + "Failed to allocate compression buffer 0, size[%x]\n", > + VE_COMP_BUFFER_SIZE); > + goto free_src1; > + } > + > + video->comp[1].virt = dma_alloc_coherent(video->dev, > + VE_COMP_BUFFER_SIZE, > + &video->comp[1].dma, > + GFP_KERNEL); > + if (!video->comp[0].virt) { > + dev_err(video->dev, > + "Failed to allocate compression buffer 1, size[%x]\n", > + VE_COMP_BUFFER_SIZE); > + goto free_comp0; > + } > + > + video->jpeg.virt = dma_alloc_coherent(video->dev, VE_JPEG_BUFFER_SIZE, > + &video->jpeg.dma, GFP_KERNEL); > + if (!video->jpeg.virt) { > + dev_err(video->dev, > + "Failed to allocate JPEG buffer, size[%x]\n", > + VE_JPEG_BUFFER_SIZE); > + goto free_comp1; > + } > + > + if (video->fmt.pixelformat == V4L2_PIX_FMT_YUV420) > + aspeed_video_init_jpeg_table(video->jpeg.virt, true); > + else > + aspeed_video_init_jpeg_table(video->jpeg.virt, false); > + > + /* > + * Calculate the memory restrictions. Don't consider the JPEG header > + * buffer since HW doesn't need to write to it. > + */ > + video->max = max(video->srcs[0].dma + VE_SRC_BUFFER_SIZE, > + video->srcs[1].dma + VE_SRC_BUFFER_SIZE); > + video->max = max(video->max, video->comp[0].dma + VE_COMP_BUFFER_SIZE); > + video->max = max(video->max, video->comp[1].dma + VE_COMP_BUFFER_SIZE); > + > + video->min = min(video->srcs[0].dma, video->srcs[1].dma); > + video->min = min(video->min, video->comp[0].dma); > + video->min = min(video->min, video->comp[1].dma); > + > + return 0; > + > +free_comp1: > + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[1].virt, > + video->comp[1].dma); > + video->comp[1].dma = 0ULL; > + video->comp[1].virt = NULL; > + > +free_comp0: > + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[0].virt, > + video->comp[0].dma); > + video->comp[0].dma = 0ULL; > + video->comp[0].virt = NULL; > + > +free_src1: > + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[1].virt, > + video->srcs[1].dma); > + video->srcs[1].dma = 0ULL; > + video->srcs[1].virt = NULL; > + > +free_src0: > + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[0].virt, > + video->srcs[0].dma); > + video->srcs[0].dma = 0ULL; > + video->srcs[0].virt = NULL; > +err: > + return -ENOMEM; > +} > + > +static void aspeed_video_free_cma(struct aspeed_video *video) > +{ > + dma_free_coherent(video->dev, VE_JPEG_BUFFER_SIZE, video->jpeg.virt, > + video->jpeg.dma); > + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[1].virt, > + video->comp[1].dma); > + dma_free_coherent(video->dev, VE_COMP_BUFFER_SIZE, video->comp[0].virt, > + video->comp[0].dma); > + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[1].virt, > + video->srcs[1].dma); > + dma_free_coherent(video->dev, VE_SRC_BUFFER_SIZE, video->srcs[0].virt, > + video->srcs[0].dma); > + > + video->srcs[0].dma = 0ULL; > + video->srcs[0].virt = NULL; > + video->srcs[1].dma = 0ULL; > + video->srcs[1].virt = NULL; > + video->comp[0].dma = 0ULL; > + video->comp[0].virt = NULL; > + video->comp[1].dma = 0ULL; > + video->comp[1].virt = NULL; > + video->jpeg.dma = 0ULL; > + video->jpeg.virt = NULL; > +} > + > +static int aspeed_video_start(struct aspeed_video *video) > +{ > + int rc = aspeed_video_allocate_cma(video); > + > + if (rc) > + return rc; > + > + aspeed_video_on(video); > + > + aspeed_video_init_regs(video); > + > + rc = aspeed_video_get_resolution(video); > + if (rc) > + aspeed_video_free_cma(video); > + > + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); > + > + return rc; > +} > + > +static void aspeed_video_stop(struct aspeed_video *video) > +{ > + cancel_delayed_work_sync(&video->res_work); > + > + aspeed_video_off(video); > + > + aspeed_video_free_cma(video); > + > + clear_bit(VIDEO_FRAME_AVAILABLE, &video->flags); > +} > + > +static int aspeed_video_querycap(struct file *file, void *fh, > + struct v4l2_capability *cap) > +{ > + struct aspeed_video *video = video_drvdata(file); > + > + strncpy(cap->driver, DEVICE_NAME, sizeof(cap->driver)); Use strlcpy. Also fill in bus_info ("platform:<foo>"). > + cap->capabilities = video->vdev.device_caps | V4L2_CAP_DEVICE_CAPS; You can drop this, it's filled in for you. > + > + return 0; > +} > + > +static int aspeed_video_get_format(struct file *file, void *fh, > + struct v4l2_format *f) > +{ > + int rc; > + struct aspeed_video *video = video_drvdata(file); > + > + if (test_bit(VIDEO_RES_CHANGE, &video->flags)) { > + if (file->f_flags & O_NONBLOCK) > + return -EAGAIN; > + > + rc = wait_event_interruptible(video->wait, > + !test_bit(VIDEO_RES_CHANGE, > + &video->flags)); > + if (rc) > + return -EINTR; No, get_format should always just return the currently set format, not the format that is detected. Use VIDIOC_QUERY_DV_TIMINGS for that. > + } > + > + f->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; Drop this, not needed. > + f->fmt.pix = video->fmt; > + > + return 0; > +} > + > +static int aspeed_video_set_format(struct file *file, void *fh, > + struct v4l2_format *f) > +{ > + struct aspeed_video *video = video_drvdata(file); > + > + if (f->fmt.pix.pixelformat == video->fmt.pixelformat) > + return 0; > + > + if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV444) { > + video->fmt.pixelformat = V4L2_PIX_FMT_YUV444; > + aspeed_video_init_jpeg_table(video->jpeg.virt, false); > + aspeed_video_update(video, VE_SEQ_CTRL, ~VE_SEQ_CTRL_YUV420, > + 0); > + } else if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV420) { > + video->fmt.pixelformat = V4L2_PIX_FMT_YUV420; > + aspeed_video_init_jpeg_table(video->jpeg.virt, true); > + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, > + VE_SEQ_CTRL_YUV420); This isn't filling in any of the other fields. > + } else { > + return -EINVAL; > + } > + > + return 0; > +} > + > +static int aspeed_video_get_jpegcomp(struct file *file, void *fh, > + struct v4l2_jpegcompression *a) > +{ > + struct aspeed_video *video = video_drvdata(file); > + > + a->quality = video->jpeg_quality; > + > + return 0; > +} > + > +static int aspeed_video_set_jpegcomp(struct file *file, void *fh, > + const struct v4l2_jpegcompression *a) > +{ > + u32 comp_ctrl; > + struct aspeed_video *video = video_drvdata(file); > + > + if (a->quality < 0 || a->quality > 11) > + return -EINVAL; > + > + video->jpeg_quality = a->quality; > + comp_ctrl = FIELD_PREP(VE_COMP_CTRL_DCT_LUM, video->jpeg_quality) | > + FIELD_PREP(VE_COMP_CTRL_DCT_CHR, video->jpeg_quality | 0x10); > + > + aspeed_video_update(video, VE_COMP_CTRL, > + ~(VE_COMP_CTRL_DCT_LUM | VE_COMP_CTRL_DCT_CHR), > + comp_ctrl); > + > + return 0; > +} As the spec says, the jpegcomp ioctls are deprecated and you should use JPEG controls instead. See: https://hverkuil.home.xs4all.nl/spec/uapi/v4l/vidioc-g-jpegcomp.html I stop reviewing here, since the first thing you need to do is to run v4l2-compliance for your device driver. See my reply to patch 0/4. Regards, Hans > + > +static int aspeed_video_get_parm(struct file *file, void *fh, > + struct v4l2_streamparm *a) > +{ > + struct aspeed_video *video = video_drvdata(file); > + > + a->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > + a->parm.capture.timeperframe.numerator = 1; > + a->parm.capture.timeperframe.denominator = video->frame_rate; > + > + return 0; > +} > + > +static int aspeed_video_set_parm(struct file *file, void *fh, > + struct v4l2_streamparm *a) > +{ > + int frame_rate; > + struct aspeed_video *video = video_drvdata(file); > + > + if (a->type != V4L2_BUF_TYPE_VIDEO_CAPTURE) > + return -EINVAL; > + > + frame_rate = a->parm.capture.timeperframe.denominator / > + a->parm.capture.timeperframe.numerator; > + > + if (frame_rate < 0 || frame_rate > 60) > + return -EINVAL; > + > + if (video->frame_rate != frame_rate) { > + video->frame_rate = frame_rate; > + aspeed_video_update(video, VE_CTRL, ~VE_CTRL_FRC, > + FIELD_PREP(VE_CTRL_FRC, frame_rate)); > + } > + > + return 0; > +} > + > +static const struct v4l2_ioctl_ops aspeed_video_ioctls = { > + .vidioc_querycap = aspeed_video_querycap, > + .vidioc_g_fmt_vid_cap = aspeed_video_get_format, > + .vidioc_s_fmt_vid_cap = aspeed_video_set_format, > + .vidioc_g_jpegcomp = aspeed_video_get_jpegcomp, > + .vidioc_s_jpegcomp = aspeed_video_set_jpegcomp, > + .vidioc_g_parm = aspeed_video_get_parm, > + .vidioc_s_parm = aspeed_video_set_parm, > +}; > + > +static void aspeed_video_resolution_work(struct work_struct *work) > +{ > + int rc; > + struct delayed_work *dwork = to_delayed_work(work); > + struct aspeed_video *video = container_of(dwork, struct aspeed_video, > + res_work); > + > + /* No clients remaining after delay */ > + if (atomic_read(&video->clients) == 0) > + goto done; > + > + aspeed_video_on(video); > + > + aspeed_video_init_regs(video); > + > + rc = aspeed_video_get_resolution(video); > + if (rc) { > + dev_err(video->dev, > + "resolution changed; couldn't get new resolution\n"); > + } else { > + video->frame_idx = 0; > + clear_bit(VIDEO_FRAME_TRIGGERED, &video->flags); > + } > + > +done: > + clear_bit(VIDEO_RES_CHANGE, &video->flags); > + wake_up_interruptible_all(&video->wait); > +} > + > +static bool aspeed_video_frame_available(struct aspeed_video *video) > +{ > + if (!test_and_clear_bit(VIDEO_FRAME_AVAILABLE, &video->flags)) { > + if (!test_bit(VIDEO_FRAME_TRIGGERED, &video->flags)) > + aspeed_video_start_frame(video); > + > + return false; > + } > + > + return true; > +} > + > +static ssize_t aspeed_video_file_read(struct file *file, char __user *buf, > + size_t count, loff_t *ppos) > +{ > + int rc; > + int fidx; > + size_t size; > + struct aspeed_video *video = video_drvdata(file); > + > + if (mutex_lock_interruptible(&video->video_lock)) > + return -EINTR; > + > + if (file->f_flags & O_NONBLOCK) { > + if (!aspeed_video_frame_available(video)) { > + rc = -EAGAIN; > + goto unlock; > + } else { > + goto ready; > + } > + } > + > + rc = wait_event_interruptible(video->wait, > + aspeed_video_frame_available(video)); > + if (rc) { > + rc = -EINTR; > + goto unlock; > + } > + > +ready: > + fidx = video->frame_idx; > + size = min_t(size_t, video->frame_size, count); > + aspeed_video_start_frame(video); > + > + if (copy_to_user(buf, video->comp[fidx].virt, size)) { > + rc = -EFAULT; > + goto unlock; > + } > + > + rc = size; > + > +unlock: > + mutex_unlock(&video->video_lock); > + return rc; > +} > + > +static int aspeed_video_open(struct file *file) > +{ > + int rc; > + struct aspeed_video *video = video_drvdata(file); > + > + if (atomic_inc_return(&video->clients) == 1) { > + rc = aspeed_video_start(video); > + if (rc) { > + dev_err(video->dev, "Failed to start video engine\n"); > + atomic_dec(&video->clients); > + return rc; > + } > + } > + > + return v4l2_fh_open(file); > +} > + > +static int aspeed_video_release(struct file *file) > +{ > + int rc; > + struct aspeed_video *video = video_drvdata(file); > + > + rc = v4l2_fh_release(file); > + > + if (atomic_dec_return(&video->clients) == 0) > + aspeed_video_stop(video); > + > + return rc; > +} > + > +static const struct v4l2_file_operations aspeed_video_v4l2_fops = { > + .owner = THIS_MODULE, > + .read = aspeed_video_file_read, > + .unlocked_ioctl = video_ioctl2, > + .open = aspeed_video_open, > + .release = aspeed_video_release, > +}; > + > +static void aspeed_video_device_release(struct video_device *vdev) > +{ > +} > + > +static int aspeed_video_setup_video(struct aspeed_video *video) > +{ > + int rc; > + struct v4l2_device *v4l2_dev = &video->v4l2_dev; > + struct video_device *vdev = &video->vdev; > + > + rc = v4l2_device_register(video->dev, v4l2_dev); > + if (rc) { > + dev_err(video->dev, "Failed to register v4l2 device\n"); > + return rc; > + } > + > + vdev->fops = &aspeed_video_v4l2_fops; > + vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_READWRITE; > + vdev->v4l2_dev = v4l2_dev; > + strncpy(vdev->name, DEVICE_NAME, sizeof(vdev->name)); > + vdev->vfl_type = VFL_TYPE_GRABBER; > + vdev->vfl_dir = VFL_DIR_RX; > + vdev->release = aspeed_video_device_release; > + vdev->ioctl_ops = &aspeed_video_ioctls; > + vdev->lock = &video->video_lock; > + > + video_set_drvdata(vdev, video); > + rc = video_register_device(vdev, VFL_TYPE_GRABBER, 0); > + if (rc) { > + v4l2_device_unregister(v4l2_dev); > + dev_err(video->dev, "Failed to register video device\n"); > + return rc; > + } > + > + /* set pixel format defaults */ > + video->fmt.pixelformat = V4L2_PIX_FMT_YUV444; > + video->fmt.field = V4L2_FIELD_NONE; > + > + return 0; > +} > + > +static int aspeed_video_init(struct aspeed_video *video) > +{ > + int irq; > + int rc; > + struct device *dev = video->dev; > + > + irq = irq_of_parse_and_map(dev->of_node, 0); > + if (!irq) { > + dev_err(dev, "Unable to find IRQ\n"); > + return -ENODEV; > + } > + > + rc = devm_request_irq(dev, irq, aspeed_video_irq, IRQF_SHARED, > + DEVICE_NAME, video); > + if (rc < 0) { > + dev_err(dev, "Unable to request IRQ %d\n", irq); > + return rc; > + } > + > + video->eclk = devm_clk_get(dev, "eclk-gate"); > + if (IS_ERR(video->eclk)) { > + dev_err(dev, "Unable to get ECLK\n"); > + return PTR_ERR(video->eclk); > + } > + > + video->vclk = devm_clk_get(dev, "vclk-gate"); > + if (IS_ERR(video->vclk)) { > + dev_err(dev, "Unable to get VCLK\n"); > + return PTR_ERR(video->vclk); > + } > + > + video->rst = devm_reset_control_get_exclusive(dev, NULL); > + if (IS_ERR(video->rst)) { > + dev_err(dev, "Unable to get VE reset\n"); > + return PTR_ERR(video->rst); > + } > + > + rc = of_reserved_mem_device_init(dev); > + if (rc) { > + dev_err(dev, "Unable to reserve memory\n"); > + return rc; > + } > + > + rc = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32)); > + if (rc) { > + dev_err(dev, "Failed to set DMA mask\n"); > + of_reserved_mem_device_release(dev); > + return rc; > + } > + > + return 0; > +} > + > +static int aspeed_video_probe(struct platform_device *pdev) > +{ > + int rc; > + struct resource *res; > + struct aspeed_video *video = kzalloc(sizeof(*video), GFP_KERNEL); > + > + if (!video) > + return -ENOMEM; > + > + video->frame_rate = 30; > + video->dev = &pdev->dev; > + mutex_init(&video->video_lock); > + init_waitqueue_head(&video->wait); > + INIT_DELAYED_WORK(&video->res_work, aspeed_video_resolution_work); > + > + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); > + > + video->base = devm_ioremap_resource(video->dev, res); > + > + if (IS_ERR(video->base)) > + return PTR_ERR(video->base); > + > + rc = aspeed_video_init(video); > + if (rc) > + return rc; > + > + rc = aspeed_video_setup_video(video); > + if (rc) > + return rc; > + > + return 0; > +} > + > +static int aspeed_video_remove(struct platform_device *pdev) > +{ > + struct device *dev = &pdev->dev; > + struct v4l2_device *v4l2_dev = dev_get_drvdata(dev); > + struct aspeed_video *video = to_aspeed_video(v4l2_dev); > + > + video_unregister_device(&video->vdev); > + > + v4l2_device_unregister(v4l2_dev); > + > + of_reserved_mem_device_release(dev); > + > + return 0; > +} > + > +static const struct of_device_id aspeed_video_of_match[] = { > + { .compatible = "aspeed,ast2400-video" }, > + { .compatible = "aspeed,ast2500-video" }, > + {} > +}; > +MODULE_DEVICE_TABLE(of, aspeed_video_of_match); > + > +static struct platform_driver aspeed_video_driver = { > + .driver = { > + .name = DEVICE_NAME, > + .of_match_table = aspeed_video_of_match, > + }, > + .probe = aspeed_video_probe, > + .remove = aspeed_video_remove, > +}; > + > +module_platform_driver(aspeed_video_driver); > + > +MODULE_DESCRIPTION("ASPEED Video Engine Driver"); > +MODULE_AUTHOR("Eddie James"); > +MODULE_LICENSE("GPL v2"); > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 4/4] media: platform: Add Aspeed Video Engine driver 2018-09-03 11:40 ` Hans Verkuil @ 2018-09-13 19:00 ` Eddie James 0 siblings, 0 replies; 21+ messages in thread From: Eddie James @ 2018-09-13 19:00 UTC (permalink / raw) To: linux-arm-kernel On 09/03/2018 06:40 AM, Hans Verkuil wrote: > On 08/29/2018 11:09 PM, Eddie James wrote: >> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >> can capture and compress video data from digital or analog sources. With >> the Aspeed chip acting a service processor, the Video Engine can capture >> the host processor graphics output. >> >> Add a V4L2 driver to capture video data and compress it to JPEG images, >> making the data available through a standard read interface. >> >> Signed-off-by: Eddie James <eajames@linux.vnet.ibm.com> >> --- >> drivers/media/platform/Kconfig | 8 + >> drivers/media/platform/Makefile | 1 + >> drivers/media/platform/aspeed-video.c | 1307 +++++++++++++++++++++++++++++++++ > Missing MAINTAINERS file update. > >> 3 files changed, 1316 insertions(+) >> create mode 100644 drivers/media/platform/aspeed-video.c >> + >> +static void aspeed_video_init_jpeg_table(u32 *table, bool yuv420) >> +{ >> + int i; >> + unsigned int base; >> + >> + for (i = 0; i < ASPEED_VIDEO_JPEG_NUM_QUALITIES; i++) { >> + int j; >> + >> + base = 256 * i; /* AST HW requires this header spacing */ >> + >> + for (j = 0; j < ASPEED_VIDEO_JPEG_HEADER_SIZE; j++) >> + table[base + j] = >> + le32_to_cpu(aspeed_video_jpeg_header[j]); > This doesn't look right. These tables are in cpu format to begin with, > so le32_to_cpu doesn't make sense. > > BTW, what is the endianness of an aspeed SoC? Hi, Yes, Aspeed SoCs are LE, it's a standard ARM core. I'll do a simple copy instead of converting endianness, since this code can really only run on an Aspeed chip anyway. > >> + >> + base += ASPEED_VIDEO_JPEG_HEADER_SIZE; >> + for (j = 0; j < ASPEED_VIDEO_JPEG_DCT_SIZE; j++) >> + table[base + j] = >> + le32_to_cpu(aspeed_video_jpeg_dct[i][j]); >> + >> + base += ASPEED_VIDEO_JPEG_DCT_SIZE; >> + for (j = 0; j < ASPEED_VIDEO_JPEG_QUANT_SIZE; j++) >> + table[base + j] = >> + le32_to_cpu(aspeed_video_jpeg_quant[j]); >> + >> + if (yuv420) >> + table[base + 2] = le32_to_cpu(0x00220103); >> + } >> +} >> + >> + >> +static int aspeed_video_querycap(struct file *file, void *fh, >> + struct v4l2_capability *cap) >> +{ >> + struct aspeed_video *video = video_drvdata(file); >> + >> + strncpy(cap->driver, DEVICE_NAME, sizeof(cap->driver)); > Use strlcpy. > > Also fill in bus_info ("platform:<foo>"). > >> + cap->capabilities = video->vdev.device_caps | V4L2_CAP_DEVICE_CAPS; > You can drop this, it's filled in for you. > >> + >> + return 0; >> +} >> + >> +static int aspeed_video_get_format(struct file *file, void *fh, >> + struct v4l2_format *f) >> +{ >> + int rc; >> + struct aspeed_video *video = video_drvdata(file); >> + >> + if (test_bit(VIDEO_RES_CHANGE, &video->flags)) { >> + if (file->f_flags & O_NONBLOCK) >> + return -EAGAIN; >> + >> + rc = wait_event_interruptible(video->wait, >> + !test_bit(VIDEO_RES_CHANGE, >> + &video->flags)); >> + if (rc) >> + return -EINTR; > No, get_format should always just return the currently set format, > not the format that is detected. > > Use VIDIOC_QUERY_DV_TIMINGS for that. > >> + } >> + >> + f->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > Drop this, not needed. > >> + f->fmt.pix = video->fmt; >> + >> + return 0; >> +} >> + >> +static int aspeed_video_set_format(struct file *file, void *fh, >> + struct v4l2_format *f) >> +{ >> + struct aspeed_video *video = video_drvdata(file); >> + >> + if (f->fmt.pix.pixelformat == video->fmt.pixelformat) >> + return 0; >> + >> + if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV444) { >> + video->fmt.pixelformat = V4L2_PIX_FMT_YUV444; >> + aspeed_video_init_jpeg_table(video->jpeg.virt, false); >> + aspeed_video_update(video, VE_SEQ_CTRL, ~VE_SEQ_CTRL_YUV420, >> + 0); >> + } else if (f->fmt.pix.pixelformat == V4L2_PIX_FMT_YUV420) { >> + video->fmt.pixelformat = V4L2_PIX_FMT_YUV420; >> + aspeed_video_init_jpeg_table(video->jpeg.virt, true); >> + aspeed_video_update(video, VE_SEQ_CTRL, 0xFFFFFFFF, >> + VE_SEQ_CTRL_YUV420); > This isn't filling in any of the other fields. > >> + } else { >> + return -EINVAL; >> + } >> + >> + return 0; >> +} >> + >> +static int aspeed_video_get_jpegcomp(struct file *file, void *fh, >> + struct v4l2_jpegcompression *a) >> +{ >> + struct aspeed_video *video = video_drvdata(file); >> + >> + a->quality = video->jpeg_quality; >> + >> + return 0; >> +} >> + >> +static int aspeed_video_set_jpegcomp(struct file *file, void *fh, >> + const struct v4l2_jpegcompression *a) >> +{ >> + u32 comp_ctrl; >> + struct aspeed_video *video = video_drvdata(file); >> + >> + if (a->quality < 0 || a->quality > 11) >> + return -EINVAL; >> + >> + video->jpeg_quality = a->quality; >> + comp_ctrl = FIELD_PREP(VE_COMP_CTRL_DCT_LUM, video->jpeg_quality) | >> + FIELD_PREP(VE_COMP_CTRL_DCT_CHR, video->jpeg_quality | 0x10); >> + >> + aspeed_video_update(video, VE_COMP_CTRL, >> + ~(VE_COMP_CTRL_DCT_LUM | VE_COMP_CTRL_DCT_CHR), >> + comp_ctrl); >> + >> + return 0; >> +} > As the spec says, the jpegcomp ioctls are deprecated and you should use > JPEG controls instead. > > See: https://hverkuil.home.xs4all.nl/spec/uapi/v4l/vidioc-g-jpegcomp.html > > I stop reviewing here, since the first thing you need to do is to run > v4l2-compliance for your device driver. > > See my reply to patch 0/4. > > Regards, > > Hans Thanks for all the comments! I have addressed these items and will push up another patch set. Thanks, Eddie > >> + >> +static int aspeed_video_get_parm(struct file *file, void *fh, >> + struct v4l2_streamparm *a) >> +{ >> + struct aspeed_video *video = video_drvdata(file); >> + >> ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James ` (3 preceding siblings ...) 2018-08-29 21:09 ` [PATCH 4/4] media: platform: Add Aspeed Video Engine driver Eddie James @ 2018-08-31 17:56 ` Stephen Boyd 2018-08-31 19:30 ` Eddie James 2018-09-03 11:57 ` Hans Verkuil 5 siblings, 1 reply; 21+ messages in thread From: Stephen Boyd @ 2018-08-31 17:56 UTC (permalink / raw) To: linux-arm-kernel Quoting Eddie James (2018-08-29 14:09:29) > The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs > can capture and compress video data from digital or analog sources. With > the Aspeed chip acting as a service processor, the Video Engine can > capture the host processor graphics output. > > This series adds a V4L2 driver for the VE, providing a read() interface > only. The driver triggers the hardware to capture the host graphics output > and compress it to JPEG format. > > Testing on an AST2500 determined that the videobuf/streaming/mmap interface > was significantly slower than the simple read() interface, so I have not > included the streaming part. > > It's also possible to use an automatic mode for the VE such that > re-triggering the HW every frame isn't necessary. However this wasn't > reliable on the AST2400, and probably used more CPU anyway due to excessive > interrupts. It was approximately 15% faster. > > The series also adds the necessary parent clock definitions to the Aspeed > clock driver, with both a mux and clock divider. Please let me know your merge strategy here. I can ack the clk patches because they look fine from high-level clk driver perspective (maybe Joel can take a closer look) or I can merge the patches into clk-next and get them into next release while the video driver gets reviewed. ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-08-31 17:56 ` [PATCH 0/4] " Stephen Boyd @ 2018-08-31 19:30 ` Eddie James 2018-09-01 2:46 ` Stephen Boyd 0 siblings, 1 reply; 21+ messages in thread From: Eddie James @ 2018-08-31 19:30 UTC (permalink / raw) To: linux-arm-kernel On 08/31/2018 12:56 PM, Stephen Boyd wrote: > Quoting Eddie James (2018-08-29 14:09:29) >> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >> can capture and compress video data from digital or analog sources. With >> the Aspeed chip acting as a service processor, the Video Engine can >> capture the host processor graphics output. >> >> This series adds a V4L2 driver for the VE, providing a read() interface >> only. The driver triggers the hardware to capture the host graphics output >> and compress it to JPEG format. >> >> Testing on an AST2500 determined that the videobuf/streaming/mmap interface >> was significantly slower than the simple read() interface, so I have not >> included the streaming part. >> >> It's also possible to use an automatic mode for the VE such that >> re-triggering the HW every frame isn't necessary. However this wasn't >> reliable on the AST2400, and probably used more CPU anyway due to excessive >> interrupts. It was approximately 15% faster. >> >> The series also adds the necessary parent clock definitions to the Aspeed >> clock driver, with both a mux and clock divider. > Please let me know your merge strategy here. I can ack the clk patches > because they look fine from high-level clk driver perspective (maybe > Joel can take a closer look) or I can merge the patches into clk-next > and get them into next release while the video driver gets reviewed. Thanks for taking a look! Probably preferable to get the clk patches into clk-next first (though Joel reviewing would be great). I just put everything in the same set for the sake of explaining the necessity of the clk changes. Thanks, Eddie > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-08-31 19:30 ` Eddie James @ 2018-09-01 2:46 ` Stephen Boyd 0 siblings, 0 replies; 21+ messages in thread From: Stephen Boyd @ 2018-09-01 2:46 UTC (permalink / raw) To: linux-arm-kernel Quoting Eddie James (2018-08-31 12:30:02) > > > On 08/31/2018 12:56 PM, Stephen Boyd wrote: > > Quoting Eddie James (2018-08-29 14:09:29) > > Please let me know your merge strategy here. I can ack the clk patches > > because they look fine from high-level clk driver perspective (maybe > > Joel can take a closer look) or I can merge the patches into clk-next > > and get them into next release while the video driver gets reviewed. > > Thanks for taking a look! Probably preferable to get the clk patches > into clk-next first (though Joel reviewing would be great). I just put > everything in the same set for the sake of explaining the necessity of > the clk changes. > Ok. If Joel is able to review then I can easily merge it into clk-next. ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James ` (4 preceding siblings ...) 2018-08-31 17:56 ` [PATCH 0/4] " Stephen Boyd @ 2018-09-03 11:57 ` Hans Verkuil 2018-09-13 19:11 ` Eddie James 5 siblings, 1 reply; 21+ messages in thread From: Hans Verkuil @ 2018-09-03 11:57 UTC (permalink / raw) To: linux-arm-kernel Hi Eddie, Thank you for your work on this. Interesting to see support for this SoC :-) On 08/29/2018 11:09 PM, Eddie James wrote: > The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs > can capture and compress video data from digital or analog sources. With > the Aspeed chip acting as a service processor, the Video Engine can > capture the host processor graphics output. > > This series adds a V4L2 driver for the VE, providing a read() interface > only. The driver triggers the hardware to capture the host graphics output > and compress it to JPEG format. > > Testing on an AST2500 determined that the videobuf/streaming/mmap interface > was significantly slower than the simple read() interface, so I have not > included the streaming part. Do you know why? It should be equal or faster, not slower. I reviewed about half of the driver, but then I stopped since there were too many things missing. First of all, you need to test your driver with v4l2-compliance (available here: https://git.linuxtv.org/v4l-utils.git/). Always compile from the git repo since the versions from distros tend to be too old. Just run 'v4l2-compliance -d /dev/videoX' and fix all issues. Then run 'v4l2-compliance -s -d /dev/videoX' to test streaming. This utility checks if the driver follows the V4L2 API correctly, implements all ioctls that it should and fills in all the fields that it should. Please add the output of 'v4l2-compliance -s' to future versions of this patch series: I don't accept V4L2 drivers without a clean report of this utility. If you have any questions, then mail me or (usually quicker) ask on the #v4l freenode irc channel (I'm in the CET timezone). One thing that needs more explanation: from what I could tell from the driver the VIDIOC_G_FMT ioctl returns the detected format instead of the current format. This is wrong. Instead you should implement the VIDIOC_*_DV_TIMINGS ioctls and the V4L2_EVENT_SOURCE_CHANGE event. The normal sequence is that userspace queries the current timings with VIDIOC_QUERY_DV_TIMINGS, if it finds valid timings, then it sets these timings with _S_DV_TIMINGS. Now it can call G/S_FMT. If the timings change, then the driver should detect that and send a V4L2_EVENT_SOURCE_CHANGE event. When the application receives this event it can take action, such as increasing the size of the buffer for the jpeg data that it reads into. The reason for this sequence of events is that you can't just change the format/resolution mid-stream without giving userspace the chance to reconfigure. Regards, Hans > > It's also possible to use an automatic mode for the VE such that > re-triggering the HW every frame isn't necessary. However this wasn't > reliable on the AST2400, and probably used more CPU anyway due to excessive > interrupts. It was approximately 15% faster. > > The series also adds the necessary parent clock definitions to the Aspeed > clock driver, with both a mux and clock divider. > > Eddie James (4): > clock: aspeed: Add VIDEO reset index definition > clock: aspeed: Setup video engine clocking > dt-bindings: media: Add Aspeed Video Engine binding documentation > media: platform: Add Aspeed Video Engine driver > > .../devicetree/bindings/media/aspeed-video.txt | 23 + > drivers/clk/clk-aspeed.c | 41 +- > drivers/media/platform/Kconfig | 8 + > drivers/media/platform/Makefile | 1 + > drivers/media/platform/aspeed-video.c | 1307 ++++++++++++++++++++ > include/dt-bindings/clock/aspeed-clock.h | 1 + > 6 files changed, 1379 insertions(+), 2 deletions(-) > create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt > create mode 100644 drivers/media/platform/aspeed-video.c > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-09-03 11:57 ` Hans Verkuil @ 2018-09-13 19:11 ` Eddie James 2018-09-14 6:56 ` Hans Verkuil 0 siblings, 1 reply; 21+ messages in thread From: Eddie James @ 2018-09-13 19:11 UTC (permalink / raw) To: linux-arm-kernel On 09/03/2018 06:57 AM, Hans Verkuil wrote: > Hi Eddie, > > Thank you for your work on this. Interesting to see support for this SoC :-) > > On 08/29/2018 11:09 PM, Eddie James wrote: >> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >> can capture and compress video data from digital or analog sources. With >> the Aspeed chip acting as a service processor, the Video Engine can >> capture the host processor graphics output. >> >> This series adds a V4L2 driver for the VE, providing a read() interface >> only. The driver triggers the hardware to capture the host graphics output >> and compress it to JPEG format. >> >> Testing on an AST2500 determined that the videobuf/streaming/mmap interface >> was significantly slower than the simple read() interface, so I have not >> included the streaming part. > Do you know why? It should be equal or faster, not slower. Yes, it seems to be an issue with the timing of the video engine interrupts compared with how a normal v4l2 application queues buffers. With the simple read() application, the driver can swap between DMA buffers freely and get a frame ahead. With the streaming buffers, I found the driver ran through the queue quite quickly, but then, once userspace queues again, we had to wait for the next frame, as I couldn't get a frame ahead since no buffers were available during that time period. This could possibly be solved with more buffers but this gets to require a lot of memory, since each buffer is allocated for the full frame size even though we only fill a fraction of it with JPEG data... > > I reviewed about half of the driver, but then I stopped since there were too > many things missing. > > First of all, you need to test your driver with v4l2-compliance (available here: > https://git.linuxtv.org/v4l-utils.git/). Always compile from the git repo since > the versions from distros tend to be too old. > > Just run 'v4l2-compliance -d /dev/videoX' and fix all issues. Then run > 'v4l2-compliance -s -d /dev/videoX' to test streaming. > > This utility checks if the driver follows the V4L2 API correctly, implements > all ioctls that it should and fills in all the fields that it should. > > Please add the output of 'v4l2-compliance -s' to future versions of this patch > series: I don't accept V4L2 drivers without a clean report of this utility. Sure thing. Thanks for the guidance. > > If you have any questions, then mail me or (usually quicker) ask on the #v4l > freenode irc channel (I'm in the CET timezone). > > One thing that needs more explanation: from what I could tell from the driver > the VIDIOC_G_FMT ioctl returns the detected format instead of the current > format. This is wrong. Instead you should implement the VIDIOC_*_DV_TIMINGS > ioctls and the V4L2_EVENT_SOURCE_CHANGE event. > > The normal sequence is that userspace queries the current timings with > VIDIOC_QUERY_DV_TIMINGS, if it finds valid timings, then it sets these > timings with _S_DV_TIMINGS. Now it can call G/S_FMT. If the timings > change, then the driver should detect that and send a V4L2_EVENT_SOURCE_CHANGE > event. OK I see. I ended up simplifying this part anyway since it's not possible to change the video size from the driver. I don't think there is a need for VIDIOC_QUERY_DV_TIMINGS now, but feel free to review. Thanks again, Eddie > > When the application receives this event it can take action, such as > increasing the size of the buffer for the jpeg data that it reads into. > > The reason for this sequence of events is that you can't just change the > format/resolution mid-stream without giving userspace the chance to > reconfigure. > > Regards, > > Hans > >> It's also possible to use an automatic mode for the VE such that >> re-triggering the HW every frame isn't necessary. However this wasn't >> reliable on the AST2400, and probably used more CPU anyway due to excessive >> interrupts. It was approximately 15% faster. >> >> The series also adds the necessary parent clock definitions to the Aspeed >> clock driver, with both a mux and clock divider. >> >> Eddie James (4): >> clock: aspeed: Add VIDEO reset index definition >> clock: aspeed: Setup video engine clocking >> dt-bindings: media: Add Aspeed Video Engine binding documentation >> media: platform: Add Aspeed Video Engine driver >> >> .../devicetree/bindings/media/aspeed-video.txt | 23 + >> drivers/clk/clk-aspeed.c | 41 +- >> drivers/media/platform/Kconfig | 8 + >> drivers/media/platform/Makefile | 1 + >> drivers/media/platform/aspeed-video.c | 1307 ++++++++++++++++++++ >> include/dt-bindings/clock/aspeed-clock.h | 1 + >> 6 files changed, 1379 insertions(+), 2 deletions(-) >> create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt >> create mode 100644 drivers/media/platform/aspeed-video.c >> ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-09-13 19:11 ` Eddie James @ 2018-09-14 6:56 ` Hans Verkuil 2018-09-14 15:07 ` Eddie James 0 siblings, 1 reply; 21+ messages in thread From: Hans Verkuil @ 2018-09-14 6:56 UTC (permalink / raw) To: linux-arm-kernel On 09/13/2018 09:11 PM, Eddie James wrote: > > > On 09/03/2018 06:57 AM, Hans Verkuil wrote: >> Hi Eddie, >> >> Thank you for your work on this. Interesting to see support for this SoC :-) >> >> On 08/29/2018 11:09 PM, Eddie James wrote: >>> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >>> can capture and compress video data from digital or analog sources. With >>> the Aspeed chip acting as a service processor, the Video Engine can >>> capture the host processor graphics output. >>> >>> This series adds a V4L2 driver for the VE, providing a read() interface >>> only. The driver triggers the hardware to capture the host graphics output >>> and compress it to JPEG format. >>> >>> Testing on an AST2500 determined that the videobuf/streaming/mmap interface >>> was significantly slower than the simple read() interface, so I have not >>> included the streaming part. >> Do you know why? It should be equal or faster, not slower. > > Yes, it seems to be an issue with the timing of the video engine > interrupts compared with how a normal v4l2 application queues buffers. > With the simple read() application, the driver can swap between DMA > buffers freely and get a frame ahead. With the streaming buffers, I > found the driver ran through the queue quite quickly, but then, once > userspace queues again, we had to wait for the next frame, as I couldn't > get a frame ahead since no buffers were available during that time > period. This could possibly be solved with more buffers but this gets to > require a lot of memory, since each buffer is allocated for the full > frame size even though we only fill a fraction of it with JPEG data... For stream I/O you usually need 3 buffers: one is being DMAed, one is the next queued up frame for the DMA and the third is being processed in userspace. If userspace doesn't process buffers fast enough, then the driver will need to capture into the same buffer over and over again until userspace finally queues another buffer. What I don't understand here is what the frame rate is. Is the capture framerate the same as the host processor graphics output? Or is it unrelated to that? The problem with read() is that 1) it requires copying the video data, and 2) you cannot use dmabuf for zero-copying pipelines. Whether or not 2 is needed depends on your hardware. > >> >> I reviewed about half of the driver, but then I stopped since there were too >> many things missing. >> >> First of all, you need to test your driver with v4l2-compliance (available here: >> https://git.linuxtv.org/v4l-utils.git/). Always compile from the git repo since >> the versions from distros tend to be too old. >> >> Just run 'v4l2-compliance -d /dev/videoX' and fix all issues. Then run >> 'v4l2-compliance -s -d /dev/videoX' to test streaming. >> >> This utility checks if the driver follows the V4L2 API correctly, implements >> all ioctls that it should and fills in all the fields that it should. >> >> Please add the output of 'v4l2-compliance -s' to future versions of this patch >> series: I don't accept V4L2 drivers without a clean report of this utility. > > Sure thing. Thanks for the guidance. > >> >> If you have any questions, then mail me or (usually quicker) ask on the #v4l >> freenode irc channel (I'm in the CET timezone). >> >> One thing that needs more explanation: from what I could tell from the driver >> the VIDIOC_G_FMT ioctl returns the detected format instead of the current >> format. This is wrong. Instead you should implement the VIDIOC_*_DV_TIMINGS >> ioctls and the V4L2_EVENT_SOURCE_CHANGE event. >> >> The normal sequence is that userspace queries the current timings with >> VIDIOC_QUERY_DV_TIMINGS, if it finds valid timings, then it sets these >> timings with _S_DV_TIMINGS. Now it can call G/S_FMT. If the timings >> change, then the driver should detect that and send a V4L2_EVENT_SOURCE_CHANGE >> event. > > OK I see. I ended up simplifying this part anyway since it's not > possible to change the video size from the driver. I don't think there > is a need for VIDIOC_QUERY_DV_TIMINGS now, but feel free to review. You are capturing the host graphics output, right? So you need to know the resolution, timings, etc. of that output? What if the host changes resolution? That's something you need to know, or am I missing something? This is obviously a somewhat different environment than what I am used to, so bear with me if I ask stupid questions... Regards, Hans > > Thanks again, > Eddie > >> >> When the application receives this event it can take action, such as >> increasing the size of the buffer for the jpeg data that it reads into. >> >> The reason for this sequence of events is that you can't just change the >> format/resolution mid-stream without giving userspace the chance to >> reconfigure. >> >> Regards, >> >> Hans >> >>> It's also possible to use an automatic mode for the VE such that >>> re-triggering the HW every frame isn't necessary. However this wasn't >>> reliable on the AST2400, and probably used more CPU anyway due to excessive >>> interrupts. It was approximately 15% faster. >>> >>> The series also adds the necessary parent clock definitions to the Aspeed >>> clock driver, with both a mux and clock divider. >>> >>> Eddie James (4): >>> clock: aspeed: Add VIDEO reset index definition >>> clock: aspeed: Setup video engine clocking >>> dt-bindings: media: Add Aspeed Video Engine binding documentation >>> media: platform: Add Aspeed Video Engine driver >>> >>> .../devicetree/bindings/media/aspeed-video.txt | 23 + >>> drivers/clk/clk-aspeed.c | 41 +- >>> drivers/media/platform/Kconfig | 8 + >>> drivers/media/platform/Makefile | 1 + >>> drivers/media/platform/aspeed-video.c | 1307 ++++++++++++++++++++ >>> include/dt-bindings/clock/aspeed-clock.h | 1 + >>> 6 files changed, 1379 insertions(+), 2 deletions(-) >>> create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt >>> create mode 100644 drivers/media/platform/aspeed-video.c >>> > ^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 0/4] media: platform: Add Aspeed Video Engine driver 2018-09-14 6:56 ` Hans Verkuil @ 2018-09-14 15:07 ` Eddie James 0 siblings, 0 replies; 21+ messages in thread From: Eddie James @ 2018-09-14 15:07 UTC (permalink / raw) To: linux-arm-kernel On 09/14/2018 01:56 AM, Hans Verkuil wrote: > On 09/13/2018 09:11 PM, Eddie James wrote: >> >> On 09/03/2018 06:57 AM, Hans Verkuil wrote: >>> Hi Eddie, >>> >>> Thank you for your work on this. Interesting to see support for this SoC :-) >>> >>> On 08/29/2018 11:09 PM, Eddie James wrote: >>>> The Video Engine (VE) embedded in the Aspeed AST2400 and AST2500 SOCs >>>> can capture and compress video data from digital or analog sources. With >>>> the Aspeed chip acting as a service processor, the Video Engine can >>>> capture the host processor graphics output. >>>> >>>> This series adds a V4L2 driver for the VE, providing a read() interface >>>> only. The driver triggers the hardware to capture the host graphics output >>>> and compress it to JPEG format. >>>> >>>> Testing on an AST2500 determined that the videobuf/streaming/mmap interface >>>> was significantly slower than the simple read() interface, so I have not >>>> included the streaming part. >>> Do you know why? It should be equal or faster, not slower. >> Yes, it seems to be an issue with the timing of the video engine >> interrupts compared with how a normal v4l2 application queues buffers. >> With the simple read() application, the driver can swap between DMA >> buffers freely and get a frame ahead. With the streaming buffers, I >> found the driver ran through the queue quite quickly, but then, once >> userspace queues again, we had to wait for the next frame, as I couldn't >> get a frame ahead since no buffers were available during that time >> period. This could possibly be solved with more buffers but this gets to >> require a lot of memory, since each buffer is allocated for the full >> frame size even though we only fill a fraction of it with JPEG data... > For stream I/O you usually need 3 buffers: one is being DMAed, one is > the next queued up frame for the DMA and the third is being processed > in userspace. If userspace doesn't process buffers fast enough, then > the driver will need to capture into the same buffer over and over again > until userspace finally queues another buffer. > > What I don't understand here is what the frame rate is. Is the capture > framerate the same as the host processor graphics output? Or is it > unrelated to that? Yes, the frame rate is the capture frame rate, unrelated to the host graphics output frame rate. It's really just a bandwidth-saving measure. > > The problem with read() is that 1) it requires copying the video data, and > 2) you cannot use dmabuf for zero-copying pipelines. Whether or not 2 is > needed depends on your hardware. > >>> I reviewed about half of the driver, but then I stopped since there were too >>> many things missing. >>> >>> First of all, you need to test your driver with v4l2-compliance (available here: >>> https://git.linuxtv.org/v4l-utils.git/). Always compile from the git repo since >>> the versions from distros tend to be too old. >>> >>> Just run 'v4l2-compliance -d /dev/videoX' and fix all issues. Then run >>> 'v4l2-compliance -s -d /dev/videoX' to test streaming. >>> >>> This utility checks if the driver follows the V4L2 API correctly, implements >>> all ioctls that it should and fills in all the fields that it should. >>> >>> Please add the output of 'v4l2-compliance -s' to future versions of this patch >>> series: I don't accept V4L2 drivers without a clean report of this utility. >> Sure thing. Thanks for the guidance. >> >>> If you have any questions, then mail me or (usually quicker) ask on the #v4l >>> freenode irc channel (I'm in the CET timezone). >>> >>> One thing that needs more explanation: from what I could tell from the driver >>> the VIDIOC_G_FMT ioctl returns the detected format instead of the current >>> format. This is wrong. Instead you should implement the VIDIOC_*_DV_TIMINGS >>> ioctls and the V4L2_EVENT_SOURCE_CHANGE event. >>> >>> The normal sequence is that userspace queries the current timings with >>> VIDIOC_QUERY_DV_TIMINGS, if it finds valid timings, then it sets these >>> timings with _S_DV_TIMINGS. Now it can call G/S_FMT. If the timings >>> change, then the driver should detect that and send a V4L2_EVENT_SOURCE_CHANGE >>> event. >> OK I see. I ended up simplifying this part anyway since it's not >> possible to change the video size from the driver. I don't think there >> is a need for VIDIOC_QUERY_DV_TIMINGS now, but feel free to review. > You are capturing the host graphics output, right? So you need to know > the resolution, timings, etc. of that output? What if the host changes > resolution? That's something you need to know, or am I missing something? Yes, that's why I had those extra lines in set_format, so that if the set format resolution matches the actual frame size, then it can change... I guess I'm probably mis-using the API. Will look into the source event change and timings calls. > > This is obviously a somewhat different environment than what I am used to, > so bear with me if I ask stupid questions... No problem! Thanks, Eddie > > Regards, > > Hans > >> Thanks again, >> Eddie >> >>> When the application receives this event it can take action, such as >>> increasing the size of the buffer for the jpeg data that it reads into. >>> >>> The reason for this sequence of events is that you can't just change the >>> format/resolution mid-stream without giving userspace the chance to >>> reconfigure. >>> >>> Regards, >>> >>> Hans >>> >>>> It's also possible to use an automatic mode for the VE such that >>>> re-triggering the HW every frame isn't necessary. However this wasn't >>>> reliable on the AST2400, and probably used more CPU anyway due to excessive >>>> interrupts. It was approximately 15% faster. >>>> >>>> The series also adds the necessary parent clock definitions to the Aspeed >>>> clock driver, with both a mux and clock divider. >>>> >>>> Eddie James (4): >>>> clock: aspeed: Add VIDEO reset index definition >>>> clock: aspeed: Setup video engine clocking >>>> dt-bindings: media: Add Aspeed Video Engine binding documentation >>>> media: platform: Add Aspeed Video Engine driver >>>> >>>> .../devicetree/bindings/media/aspeed-video.txt | 23 + >>>> drivers/clk/clk-aspeed.c | 41 +- >>>> drivers/media/platform/Kconfig | 8 + >>>> drivers/media/platform/Makefile | 1 + >>>> drivers/media/platform/aspeed-video.c | 1307 ++++++++++++++++++++ >>>> include/dt-bindings/clock/aspeed-clock.h | 1 + >>>> 6 files changed, 1379 insertions(+), 2 deletions(-) >>>> create mode 100644 Documentation/devicetree/bindings/media/aspeed-video.txt >>>> create mode 100644 drivers/media/platform/aspeed-video.c >>>> ^ permalink raw reply [flat|nested] 21+ messages in thread
end of thread, other threads:[~2018-09-14 15:07 UTC | newest] Thread overview: 21+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2018-08-29 21:09 [PATCH 0/4] media: platform: Add Aspeed Video Engine driver Eddie James 2018-08-29 21:09 ` [PATCH 1/4] clock: aspeed: Add VIDEO reset index definition Eddie James 2018-08-31 17:56 ` Stephen Boyd 2018-09-04 13:36 ` Rob Herring 2018-08-29 21:09 ` [PATCH 2/4] clock: aspeed: Setup video engine clocking Eddie James 2018-08-31 17:56 ` Stephen Boyd 2018-08-31 21:33 ` Joel Stanley 2018-08-29 21:09 ` [PATCH 3/4] dt-bindings: media: Add Aspeed Video Engine binding documentation Eddie James 2018-09-04 0:50 ` Rob Herring 2018-08-29 21:09 ` [PATCH 4/4] media: platform: Add Aspeed Video Engine driver Eddie James 2018-08-30 0:52 ` Ezequiel Garcia 2018-08-30 15:40 ` Eddie James 2018-09-03 11:40 ` Hans Verkuil 2018-09-13 19:00 ` Eddie James 2018-08-31 17:56 ` [PATCH 0/4] " Stephen Boyd 2018-08-31 19:30 ` Eddie James 2018-09-01 2:46 ` Stephen Boyd 2018-09-03 11:57 ` Hans Verkuil 2018-09-13 19:11 ` Eddie James 2018-09-14 6:56 ` Hans Verkuil 2018-09-14 15:07 ` Eddie James
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).