Re: [PATCH v4 1/7] scsi: ufs: core: Add UFSHCI_QUIRK_DME_RESET_ENABLE_AFTER_HCE

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Mani,

在 2024/11/7 23:51, Manivannan Sadhasivam 写道:
On Mon, Nov 04, 2024 at 03:31:55PM +0800, Shawn Lin wrote:
HCE on Rockchip SoC is different from both of ufshcd_hba_execute_hce()
and UFSHCI_QUIRK_BROKEN_HCE case. It need to do dme_reset and dme_enable
after enabling HCE. So in order not to abuse UFSHCI_QUIRK_BROKEN_HCE, add
a new quirk UFSHCI_QUIRK_DME_RESET_ENABLE_AFTER_HCE, to deal with that
limitation.

Suggested-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
Signed-off-by: Shawn Lin <shawn.lin@xxxxxxxxxxxxxx>
---

Changes in v4:
- fix typo

Changes in v3: None
Changes in v2: None

  drivers/ufs/core/ufshcd.c | 17 +++++++++++++++++
  include/ufs/ufshcd.h      |  6 ++++++
  2 files changed, 23 insertions(+)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 7cab1031..4084bf9 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -4819,6 +4819,7 @@ static int ufshcd_hba_execute_hce(struct ufs_hba *hba)
  {
  	int retry_outer = 3;
  	int retry_inner;
+	int ret;
start:
  	if (ufshcd_is_hba_active(hba))
@@ -4865,6 +4866,22 @@ static int ufshcd_hba_execute_hce(struct ufs_hba *hba)
  	/* enable UIC related interrupts */
  	ufshcd_enable_intr(hba, UFSHCD_UIC_MASK);
+ /*
+	 * Do dme_reset and dme_enable if a UFS host controller needs
+	 * this procedure to actually finish HCE.
+	 */
+	if (hba->quirks & UFSHCI_QUIRK_DME_RESET_ENABLE_AFTER_HCE) {
+		ret = ufshcd_dme_reset(hba);
+		if (!ret) {
+			ret = ufshcd_dme_enable(hba);
+			if (ret)
+				dev_err(hba->dev,
+					"Failed to do dme_enable after HCE.\n");

Don't you need to return failure for this and below error paths? Probably you
need to skip post change notification as well in the case of failure.

Oops, my bad.


+		} else {
+			dev_err(hba->dev, "Failed to do dme_reset after HCE.\n");
+		}
+	}
+
  	ufshcd_vops_hce_enable_notify(hba, POST_CHANGE);

Is it possible for you to carry out dme_reset() and dme_enable() in the post
change notifier of the rockchip glue driver? I'm trying to see if we can avoid
having the quirk which is only specific to Rockchip.


I will check and test this approach. Thanks.

- Mani

return 0;
diff --git a/include/ufs/ufshcd.h b/include/ufs/ufshcd.h
index a95282b..e939af8 100644
--- a/include/ufs/ufshcd.h
+++ b/include/ufs/ufshcd.h
@@ -685,6 +685,12 @@ enum ufshcd_quirks {
  	 * single doorbell mode.
  	 */
  	UFSHCD_QUIRK_BROKEN_LSDBS_CAP			= 1 << 25,
+
+	/*
+	 * This quirk needs to be enabled if host controller need to
+	 * do dme_reset and dme_enable after hce.
+	 */
+	UFSHCI_QUIRK_DME_RESET_ENABLE_AFTER_HCE		= 1 << 26,
  };
enum ufshcd_caps {
--
2.7.4







[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [SCSI Target Devel]     [Linux SCSI Target Infrastructure]     [Kernel Newbies]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Linux IIO]     [Samba]     [Device Mapper]

  Powered by Linux