ARM: cns3xxx: ahci: Fixup for softwreset failures with direct connected disks with CONFIG_SATA_PMP enabled.
Subsystems (3)
PATCH 1/1 ARM: cns3xxx: ahci: Fixup for softwreset failures with direct connected disks with CONFIG_SATA_PMP enabled.
From: Mac Lin <[email protected]>
With linux-2.6.37-rc3, arm/cns3xxx, ahci-platform driver.
If CONFIG_SATA_PMP is enabled, while not using multiplier and connect the disks directly to the board, the disk cannot be found due to software reset always failed.
[ 12.790000] ahci ahci.0: forcing PORTS_IMPL to 0x3 [ 12.800000] ahci ahci.0: AHCI 0001.0100 32 slots 2 ports 3 Gbps 0x3 impl platform mode [ 12.810000] ahci ahci.0: flags: ncq sntf pm led clo only pmp pio slum part ccc [ 12.850000] scsi0 : ahci_platform [ 12.860000] scsi1 : ahci_platform [ 12.870000] ata1: SATA max UDMA/133 irq_stat 0x00400040, connection status changed irq 65 [ 12.880000] ata2: SATA max UDMA/133 mmio [mem 0x83000000-0x83ffffff] port 0x180 irq 65 [ 13.240000] ata2: SATA link down (SStatus 0 SControl 300) [ 18.900000] ata1: link is slow to respond, please be patient (ready=0) [ 22.930000] ata1: softreset failed (device not ready) [ 28.940000] ata1: link is slow to respond, please be patient (ready=0) [ 32.970000] ata1: softreset failed (device not ready) [ 38.980000] ata1: link is slow to respond, please be patient (ready=0) [ 68.030000] ata1: softreset failed (device not ready) [ 68.040000] ata1: limiting SATA link speed to 1.5 Gbps [ 70.280000] ata1: SATA link down (SStatus 1 SControl 310)
While using multiplier with CONFIG_SATA_PMP enabled, or using disks directly without CONFIG_SATA_PMP have no issue.
The situation is that device is sending D2H Reg FIS, but controller is not reflecting it on any known means. It seems the controller has problem receiving D2H Reg FIS of the different PMP setting of the previous sent H2D Reg FIS.
This patch override the softreset function ahci_softreset by cns3xxx_ahci_softreset, which would retry ahci_do_softreset again with pmp=0 if pmp=15 failed.
--- arch/arm/mach-cns3xxx/cns3420vb.c | 2 +- arch/arm/mach-cns3xxx/devices.c | 86 +++++++++++++++++++++++++++---------- arch/arm/mach-cns3xxx/devices.h | 2 +- 3 files changed, 65 insertions(+), 25 deletions(-)
arch/arm/mach-cns3xxx/cns3420vb.c +1 -1
@@ -166,13 +166,13 @@ static struct platform_device *cns3420_pdevs[] __initdata = { &cns3420_nor_pdev, &cns3xxx_usb_ehci_device, &cns3xxx_usb_ohci_device, + &cns3xxx_ahci_pdev, };static void __init cns3420_init (void ) {platform_add_devices (cns3420_pdevs,ARRAY_SIZE (cns3420_pdevs)); -cns3xxx_ahci_init ();cns3xxx_sdhci_init (); pm_power_off = cns3xxx_power_off;
arch/arm/mach-cns3xxx/devices.c +63 -23
113 lines not on this page · raw message
arch/arm/mach-cns3xxx/devices.h +1 -1
@@ -14,7 +14,7 @@#ifndef __CNS3XXX_DEVICES_H_#define __CNS3XXX_DEVICES_H_ -void __init cns3xxx_ahci_init (void ); +extern struct platform_device cns3xxx_ahci_pdev;void __init cns3xxx_sdhci_init (void );#endif /* __CNS3XXX_DEVICES_H_ */
Signature · 1 line
1.7.3
PATCH 1/1 ARM: cns3xxx: ahci: Fixup for softwreset failures with direct connected disks with CONFIG_SATA_PMP enabled.
On Sun, Dec 05, 2010 at 12:43:41AM +0800, mkl0301 at gmail.com wrote:
[...]
Thanks for the patch!
There are few issues though.
Quoted 24 lines
arch/arm/mach-cns3xxx/devices.c +24
-0+static int cns3xxx_ahci_init (struct device *dev,void __iomem *addr) +{ +u32 tmp; + +DPRINTK ("ENTER\n" ); + + tmp =__raw_readl (MISC_SATA_POWER_MODE ); + tmp |=0x1 <<16 ;/* Disable SATA PHY 0 from SLUMBER Mode */ + tmp |=0x1 <<17 ;/* Disable SATA PHY 1 from SLUMBER Mode */ +__raw_writel (tmp,MISC_SATA_POWER_MODE ); + +/* Enable SATA PHY */ +cns3xxx_pwr_power_up (0x1 <<PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY0 ); +cns3xxx_pwr_power_up (0x1 <<PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY1 ); + +/* Enable SATA Clock */ +cns3xxx_pwr_clk_en (0x1 <<PM_CLK_GATE_REG_OFFSET_SATA ); + +/* De-Asscer SATA Reset */ +cns3xxx_pwr_soft_rst (CNS3XXX_PWR_SOFTWARE_RST (SATA )); + +return 0 ; +} +
[...]
Quoted 11 lines
arch/arm/mach-cns3xxx/devices.c +0
-11-void __init cns3xxx_ahci_init (void ) -{ -u32 tmp; - - tmp =__raw_readl (MISC_SATA_POWER_MODE ); - tmp |=0x1 <<16 ;/* Disable SATA PHY 0 from SLUMBER Mode */ - tmp |=0x1 <<17 ;/* Disable SATA PHY 1 from SLUMBER Mode */ -__raw_writel (tmp,MISC_SATA_POWER_MODE ); - -/* Enable SATA PHY */ -cns3xxx_pwr_power_up (0x1 <<PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY0 );
arch/arm/mach-cns3xxx/devices.c static struct platform_device cns3xxx_ahci_pdev = { · line 74
Quoted 10 lines
arch/arm/mach-cns3xxx/devices.c +0
-10-cns3xxx_pwr_power_up (0x1 <<PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY1 ); - -/* Enable SATA Clock */ -cns3xxx_pwr_clk_en (0x1 <<PM_CLK_GATE_REG_OFFSET_SATA ); - -/* De-Asscer SATA Reset */ -cns3xxx_pwr_soft_rst (CNS3XXX_PWR_SOFTWARE_RST (SATA )); - -platform_device_register (&cns3xxx_ahci_pdev); -}
This is not good because cns3xxx_pwr_*() calls aren't thread-safe. We must use them only from the "single-threaded" platform code, i.e. very early.
Once we add proper clocks (clkapi) and power management (regulators) support for CNS3xxx, we may move this into ahci->init() callback.
Plus, unfortunately this patch breaks build when AHCI_PLATFORM is set to =m.
arch/arm/mach-cns3xxx/built-in.o: In function 'cns3xxx_ahci_softreset':
cns3420vb.c:(.text+0x36c): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x390): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x3a0): undefined reference to 'ahci_check_ready' arch/arm/mach-cns3xxx/built-in.o:(.data+0x35c): undefined reference to 'ahci_ops'
make: *** [.tmp_vmlinux1] Error 1
PATCH 1/1 ARM: cns3xxx: ahci: Fixup for softwreset failures with direct connected disks with CONFIG_SATA_PMP enabled.
2010/12/21 Anton Vorontsov <[email protected]>:
Quoted 51 lines · On Sun, Dec 05, 2010 at 12:43:41AM +0800, mkl0301 at gmail.com wrote:
[...]
Thanks for the patch!
There are few issues though.
Quoted 24 lines
(quoted diff) +24
-0+static int cns3xxx_ahci_init(struct device *dev, void __iomem *addr) +{ + ? ? u32 tmp; + + ? ? DPRINTK("ENTER\n"); + + ? ? tmp = __raw_readl(MISC_SATA_POWER_MODE); + ? ? tmp |= 0x1 << 16; /* Disable SATA PHY 0 from SLUMBER Mode */ + ? ? tmp |= 0x1 << 17; /* Disable SATA PHY 1 from SLUMBER Mode */ + ? ? __raw_writel(tmp, MISC_SATA_POWER_MODE); + + ? ? /* Enable SATA PHY */ + ? ? cns3xxx_pwr_power_up(0x1 << PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY0); + ? ? cns3xxx_pwr_power_up(0x1 << PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY1); + + ? ? /* Enable SATA Clock */ + ? ? cns3xxx_pwr_clk_en(0x1 << PM_CLK_GATE_REG_OFFSET_SATA); + + ? ? /* De-Asscer SATA Reset */ + ? ? cns3xxx_pwr_soft_rst(CNS3XXX_PWR_SOFTWARE_RST(SATA)); + + ? ? return 0; +} +[...]
Quoted 19 lines
(quoted diff) +0
-19-void __init cns3xxx_ahci_init(void) -{ - ? ? u32 tmp; - - ? ? tmp = __raw_readl(MISC_SATA_POWER_MODE); - ? ? tmp |= 0x1 << 16; /* Disable SATA PHY 0 from SLUMBER Mode */ - ? ? tmp |= 0x1 << 17; /* Disable SATA PHY 1 from SLUMBER Mode */ - ? ? __raw_writel(tmp, MISC_SATA_POWER_MODE); - - ? ? /* Enable SATA PHY */ - ? ? cns3xxx_pwr_power_up(0x1 << PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY0); - ? ? cns3xxx_pwr_power_up(0x1 << PM_PLL_HM_PD_CTRL_REG_OFFSET_SATA_PHY1); - - ? ? /* Enable SATA Clock */ - ? ? cns3xxx_pwr_clk_en(0x1 << PM_CLK_GATE_REG_OFFSET_SATA); - - ? ? /* De-Asscer SATA Reset */ - ? ? cns3xxx_pwr_soft_rst(CNS3XXX_PWR_SOFTWARE_RST(SATA)); -
Quoted 8 lines
Quoted 2 lines
- ? ?
platform_device_register(&cns3xxx_ahci_pdev); -}This is not good because cns3xxx_pwr_*() calls aren't thread-safe. We must use them only from the "single-threaded" platform code, i.e. very early.
Once we add proper clocks (clkapi) and power management (regulators) support for CNS3xxx, we may move this intoahci->init()callback.
Almost every device driver needs to change those power/clk/reset bits, and it's strange to enable everything on init phase. BTW, the EHCI and OHCI that I sent previously also used these cns3xxx_pwr_*() calls.
Would they be supported in near future? If not, I would like look into those.
Quoted 8 lines
Plus, unfortunately this patch breaks build when
AHCI_PLATFORMis set to =m.arch/arm/mach-cns3xxx/built-in.o: In function 'cns3xxx_ahci_softreset':
cns3420vb.c:(.text+0x36c): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x390): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x3a0): undefined reference to 'ahci_check_ready' arch/arm/mach-cns3xxx/built-in.o:(.data+0x35c): undefined reference to 'ahci_ops'
make: *** [.tmp_vmlinux1] Error 1
Sorry that I didn't notice this. I was trying to reuse ahci-platform as possible.
Building platform device as module seems strange to me, therefore I would add a new ahci-platform alike platform driver.
Best Regards,
Mac Lin
PATCH 1/1 ARM: cns3xxx: ahci: Fixup for softwreset failures with direct connected disks with CONFIG_SATA_PMP enabled.
On Wed, Dec 22, 2010 at 05:27:20PM +0800, Lin Mac wrote:
[...]
Quoted 10 lines
Quoted 5 lines
This is not good because cns3xxx_pwr_*() calls aren't thread-safe. We must use them only from the "single-threaded" platform code, i.e. very early.
Once we add proper clocks (clkapi) and power management (regulators) support for CNS3xxx, we may move this intoahci->init()callback.Almost every device driver needs to change those power/clk/reset bits, and it's strange to enable everything on init phase. BTW, the EHCI and OHCI that I sent previously also used these cns3xxx_pwr_*() calls.
Would they be supported in near future? If not, I would like look into those.
Currently I do not work on adding these features, so feel free look into it.
Quoted 8 lines
Quoted 8 lines
Plus, unfortunately this patch breaks build when
AHCI_PLATFORMis set to =m.arch/arm/mach-cns3xxx/built-in.o: In function 'cns3xxx_ahci_softreset':
cns3420vb.c:(.text+0x36c): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x390): undefined reference to 'ahci_do_softreset' cns3420vb.c:(.text+0x3a0): undefined reference to 'ahci_check_ready' arch/arm/mach-cns3xxx/built-in.o:(.data+0x35c): undefined reference to 'ahci_ops'
make: *** [.tmp_vmlinux1] Error 1
Quoted 4 lines
Sorry that I didn't notice this. I was trying to reuse ahci-platform as possible.
Building platform device as module seems strange to me, therefore I would add a new ahci-platform alike platform driver.
Yep. But please make it similar to the SDHCI platform driver approach. I.e., see these commits:
commit 20b1597bcf4a76ccab232fa032f5f9ad30069167
Author: Anton Vorontsov <[email protected]> Date: Tue Aug 10 18:01:49 2010 -0700
sdhci-pltfm: add support for CNS3xxx SoC devices
commit 845e3f4f06f9b1d34f39601cb6b7abfb8f40653c
Author: Anton Vorontsov <[email protected]> Date: Tue Aug 10 18:01:49 2010 -0700
sdhci-pltfm: reorganize Makefile entries to support SoC devices
commit 515033f97c0b5a1bce13fa93e09704d95b44f376
Author: Anton Vorontsov <[email protected]> Date: Tue Aug 10 18:01:47 2010 -0700
sdhci-pltfm: switch to module device table matching
Thanks,
Signature · 2 lines
Anton Vorontsov Email: cbouatmailru at gmail.com