[2/2] HACK: arndale: deinit scsi before launching Linux

Message ID 1412690200-6630-2-git-send-email-ijc@hellion.org.uk
State New
Headers show

Commit Message

Ian Campbell Oct. 7, 2014, 1:56 p.m.
From: Ian Campbell <ian.campbell@citrix.com>

NOT TO BE APPLIED AS IS

Without this Linux fails to correctly init the phy (or something) and cannot
detect the disk.

Even with this we can fail to detect the disk outselves on some fraction of
boots, so something else is clearly up too.
---
 arch/arm/cpu/armv7/exynos/sata.c | 11 +++++++++++
 arch/arm/lib/bootm.c             |  4 ++++
 board/samsung/arndale/arndale.c  |  5 +++++
 drivers/block/ahci.c             |  4 ++++
 include/scsi.h                   |  1 +
 5 files changed, 25 insertions(+)

Comments

Wolfgang Denk Oct. 8, 2014, 8:18 a.m. | #1
Dear Ian Campbell,

In message <1412690200-6630-2-git-send-email-ijc@hellion.org.uk> you wrote:
> From: Ian Campbell <ian.campbell@citrix.com>
> 
> NOT TO BE APPLIED AS IS
> 
> Without this Linux fails to correctly init the phy (or something) and cannot
> detect the disk.
> 
> Even with this we can fail to detect the disk outselves on some fraction of
> boots, so something else is clearly up too.

Why does the Subject: say "scsi" when you actually mean SATA?

Note that this is a bug fix.  U-Boot should _always_ shut down all
peripherals it used before booting an OS, see [1]

[1] http://www.denx.de/wiki/U-Boot/DesignPrinciples

Best regards,

Wolfgang Denk
Ian Campbell Oct. 9, 2014, 7:40 a.m. | #2
On Wed, 2014-10-08 at 10:18 +0200, Wolfgang Denk wrote:
> Dear Ian Campbell,
> 
> In message <1412690200-6630-2-git-send-email-ijc@hellion.org.uk> you wrote:
> > From: Ian Campbell <ian.campbell@citrix.com>
> > 
> > NOT TO BE APPLIED AS IS
> > 
> > Without this Linux fails to correctly init the phy (or something) and cannot
> > detect the disk.
> > 
> > Even with this we can fail to detect the disk outselves on some fraction of
> > boots, so something else is clearly up too.
> 
> Why does the Subject: say "scsi" when you actually mean SATA?

The patch is adding "scsi_deinit" to mirror the existing "scsi_init"
function, and the u-boot CLI command (and dev name to load etc) is
"scsi". On this platform (as with several others) it happens that scsi
is backed by a sata device, that's all.

> Note that this is a bug fix.

But as noted above and in the series cover letter (more details there)
it is not a complete one. I'm hoping the maintainers might know of a
correct fix, which probably means "correct peripheral shutdown
processs".

Ian.

>   U-Boot should _always_ shut down all
> peripherals it used before booting an OS, see [1]
> 
> [1] http://www.denx.de/wiki/U-Boot/DesignPrinciples
> 
> Best regards,
> 
> Wolfgang Denk
>

Patch

diff --git a/arch/arm/cpu/armv7/exynos/sata.c b/arch/arm/cpu/armv7/exynos/sata.c
index 14d42e7..05f1372 100644
--- a/arch/arm/cpu/armv7/exynos/sata.c
+++ b/arch/arm/cpu/armv7/exynos/sata.c
@@ -310,6 +310,11 @@  static int exynos5_ahci_init(void __iomem *mmio)
 		printf("%s: already calibrated?\n", __func__);
 	}
 
+	/* Clear phy control enable. Seems to be necessary to
+	 * reinitialise on a warm reboot, at least sometimes. */
+	clrbits_le32(EXYNOS5_SATA_PHY_CONTROL, S5P_PMU_SATA_PHY_CONTROL_EN);
+	udelay(1000);
+
 	setbits_le32(EXYNOS5_SATA_PHY_CONTROL, S5P_PMU_SATA_PHY_CONTROL_EN);
 
 	__raw_writel(0, phy_ctrl + SATA_RESET);
@@ -368,3 +373,9 @@  int exynos5_sata_init(void)
 	}
 	return -ENODEV;
 }
+
+int exynos5_sata_deinit(void)
+{
+	clrbits_le32(EXYNOS5_SATA_PHY_CONTROL, S5P_PMU_SATA_PHY_CONTROL_EN);
+	return 0;
+}
diff --git a/arch/arm/lib/bootm.c b/arch/arm/lib/bootm.c
index 39fe7a1..9c8d242 100644
--- a/arch/arm/lib/bootm.c
+++ b/arch/arm/lib/bootm.c
@@ -15,6 +15,7 @@ 
 #include <common.h>
 #include <command.h>
 #include <image.h>
+#include <scsi.h>
 #include <u-boot/zlib.h>
 #include <asm/byteorder.h>
 #include <libfdt.h>
@@ -81,6 +82,9 @@  static void announce_and_cleanup(int fake)
 #ifdef CONFIG_USB_DEVICE
 	udc_disconnect();
 #endif
+#ifdef CONFIG_SCSI_AHCI_PLAT
+	scsi_deinit();
+#endif
 	cleanup_before_linux();
 }
 
diff --git a/board/samsung/arndale/arndale.c b/board/samsung/arndale/arndale.c
index 551bfce..43c9694 100644
--- a/board/samsung/arndale/arndale.c
+++ b/board/samsung/arndale/arndale.c
@@ -36,6 +36,11 @@  int scsi_init(void)
 	printf("ARNDALE SCSI INIT\n");
 	return exynos5_sata_init();
 }
+
+int scsi_deinit(void)
+{
+	return exynos5_sata_deinit();
+}
 #endif
 
 int board_init(void)
diff --git a/drivers/block/ahci.c b/drivers/block/ahci.c
index a93a8e1..a7e57be 100644
--- a/drivers/block/ahci.c
+++ b/drivers/block/ahci.c
@@ -959,6 +959,10 @@  void __weak scsi_init(void)
 {
 }
 
+void __weak scsi_deinit(void)
+{
+}
+
 #endif
 
 /*
diff --git a/include/scsi.h b/include/scsi.h
index 73de7b7..06ad192 100644
--- a/include/scsi.h
+++ b/include/scsi.h
@@ -172,6 +172,7 @@  void scsi_low_level_init(int busdevfunc);
  * functions residing inside cmd_scsi.c
  */
 void scsi_init(void);
+void scsi_deinit(void);
 void scsi_scan(int mode);
 
 /** @return the number of scsi disks */