scsi: ufs: ufs-pci: Intel UFS 4.0 HS-Gear5 and MCQ support
cover scsi: ufs: ufs-pci: Intel UFS 4.0 HS-Gear5 and MCQ support
096cd6b7adf2 ("scsi: ufs: ufs-pci: Add support for Intel Nova Lake") added the PCI ID for Intel's UFS 4.0 controller (8086:D335), reusing the existing Meteor Lake variant operations table. That table is missing two things this controller needs to work correctly at UFS 4.0
capability: a PA_INITIAL_ADAPT sequence for reliable HS-Gear5 link training, and the MCQ resource/runtime-config hooks required for ufshcd_alloc_mcq() to succeed. Without either, the controller still works, but silently falls back to Gear5 link instability or legacy single-doorbell mode respectively.
Patch 1 adds a pwr_change_notify hook that configures PA_INITIAL_ADAPT (resetting to the default PA_NO_ADAPT otherwise) before Gear5 power mode changes.
Patch 2 adds the mcq_config_resource and op_runtime_config hooks, with the OPR register offsets and stride this controller uses.
Both were validated on real NVL silicon: a 10-cycle reboot trial (5 minutes monitored each) with this exact configuration showed zero UIC-layer errors throughout, and MCQ hardware queues were confirmed carrying real doorbell traffic at the documented addresses.
sangram kumar yerra (2):
scsi: ufs: ufs-pci: Add support for Intel UFS 4.0 HS-Gear5
scsi: ufs: ufs-pci: Add MCQ support for Intel UFS 4.0 controllers
drivers/ufs/host/ufshcd-pci.c | 59 +++++++++++++++++++++++++++++++++++ 1 file changed, 59 insertions(+)
Signature · 1 line
2.34.1
PATCH 1/2 scsi: ufs: ufs-pci: Add support for Intel UFS 4.0 HS-Gear5
Reliable HS-Gear5 operation on Intel UFS 4.0 controllers requires configuring PA_INITIAL_ADAPT before changing the power mode. Without this setting, the link fails to train reliably at Gear5.
Add a pwr_change_notify() hook to configure the adaptation mode before the power mode transition. Enable this only for UFS 4.0 and later controllers by checking hba->ufs_version.
Wire the hook into the existing Meteor Lake family variant operations table (ufs_intel_mtl_hba_vops) instead of introducing a separate table, since the Intel UFS 4.0 PCI variant (PCI ID 8086:D335) already uses this vops table and the hook is internally gated on UFS version >= 4.0.
Use PA_INITIAL_ADAPT when the negotiated TX power mode is FAST_MODE or FASTAUTO_MODE. Otherwise, reset the adaptation mode to PA_NO_ADAPT, which is the default setting.
--- drivers/ufs/host/ufshcd-pci.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)
drivers/ufs/host/ufshcd-pci.c +20 -0
@@ -181,6 +181,25 @@ static int ufs_intel_lkf_pwr_change_notify(struct ufs_hba *hba,return err; } +static int ufs_intel_nvl_pwr_change_notify (struct ufs_hba *hba, +enum ufs_notify_change_status stage, +struct ufs_pa_layer_attr *dev_req_params) +{ +int adapt_val; + +if (stage !=PRE_CHANGE || hba->ufs_version <ufshci_version (4 ,0 )) +return 0 ; + +if (dev_req_params->pwr_tx ==FAST_MODE || dev_req_params->pwr_tx ==FASTAUTO_MODE ) + adapt_val =PA_INITIAL_ADAPT ; +else + adapt_val =PA_NO_ADAPT ; + +ufshcd_dme_configure_adapt (hba, dev_req_params->gear_tx, adapt_val); + +return 0 ; +} +static int ufs_intel_lkf_apply_dev_quirks (struct ufs_hba *hba) {u32 granularity, peer_granularity;
@@ -527,6 +546,7 @@ static struct ufs_hba_variant_ops ufs_intel_mtl_hba_vops = { .exit = ufs_intel_common_exit, .hce_enable_notify = ufs_intel_hce_enable_notify, .link_startup_notify = ufs_intel_link_startup_notify, + .pwr_change_notify = ufs_intel_nvl_pwr_change_notify, .resume = ufs_intel_resume, .device_reset = ufs_intel_device_reset, };
Signature · 1 line
2.34.1
PATCH 2/2 scsi: ufs: ufs-pci: Add MCQ support for Intel UFS 4.0 controllers
In reply to sangram kumar yerraThe Intel UFS 4.0 PCI variant (PCI ID 8086:D335) advertises MCQ support in its capability register. However, ufshcd_alloc_mcq() also requires an .op_runtime_config hook to locate the per-queue operation and runtime (OPR) register blocks, which was not provided by this variant operations table.
As a result, MCQ initialization fails and ufshcd_add_scsi_host() prints "MCQ mode is disabled, err=%d\n" before falling back to legacy single-doorbell (SDB) mode.
Add ufs_intel_mcq_config_resource() to initialize the MCQ configuration base and add ufs_intel_op_runtime_config() to set up the OPR register offsets and stride.
Wire both hooks into the variant operations table so MCQ is enabled when supported by the hardware.
--- drivers/ufs/host/ufshcd-pci.c | 39 +++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+)
drivers/ufs/host/ufshcd-pci.c +39 -0
@@ -460,6 +460,43 @@ static int ufs_intel_mtl_init(struct ufs_hba *hba)return ufs_intel_common_init (hba); } +static int ufs_intel_mcq_config_resource (struct ufs_hba *hba) +{ + hba->mcq_base = hba->mmio_base +ufshcd_mcq_queue_cfg_addr (hba); + +return 0 ; +} + +/* +* This Intel UFS4.0 controller maps MCQ doorbell and interrupt-status +* registers into the same PCI BAR as the legacy HCI space, at this +* fixed offset/stride. +*/ +#define UFS_INTEL_SQDAO0 0x2800 +#define UFS_INTEL_SQISAO0 0x2814 +#define UFS_INTEL_CQDAO0 0x281C +#define UFS_INTEL_CQISAO0 0x2824 +#define UFS_INTEL_MCQ_STRIDE 0x30 + +static int ufs_intel_op_runtime_config (struct ufs_hba *hba) +{ +struct ufshcd_mcq_opr_info_t *opr; +int i; + + hba->mcq_opr[OPR_SQD ].offset =UFS_INTEL_SQDAO0 ; + hba->mcq_opr[OPR_SQIS ].offset =UFS_INTEL_SQISAO0 ; + hba->mcq_opr[OPR_CQD ].offset =UFS_INTEL_CQDAO0 ; + hba->mcq_opr[OPR_CQIS ].offset =UFS_INTEL_CQISAO0 ; + +for (i =0 ; i <OPR_MAX ; i++) { + opr = &hba->mcq_opr[i]; + opr->stride =UFS_INTEL_MCQ_STRIDE ; + opr->base = hba->mmio_base + opr->offset; + } + +return 0 ; +} +static int ufs_qemu_get_hba_mac (struct ufs_hba *hba) {return MAX_SUPP_MAC ;
@@ -547,6 +584,8 @@ static struct ufs_hba_variant_ops ufs_intel_mtl_hba_vops = { .hce_enable_notify = ufs_intel_hce_enable_notify, .link_startup_notify = ufs_intel_link_startup_notify, .pwr_change_notify = ufs_intel_nvl_pwr_change_notify, + .mcq_config_resource = ufs_intel_mcq_config_resource, + .op_runtime_config = ufs_intel_op_runtime_config, .resume = ufs_intel_resume, .device_reset = ufs_intel_device_reset, };
Signature · 1 line
2.34.1
Quoted 1 line · sangram kumar yerra
[ ... ]
sangram,
Quoted 9 lines
096cd6b7adf2("scsi: ufs: ufs-pci: Add support for Intel Nova Lake") added the PCI ID for Intel's UFS 4.0 controller (8086:D335), reusing the existing Meteor Lake variant operations table. That table is missing two things this controller needs to work correctly at UFS 4.0
capability: aPA_INITIAL_ADAPTsequence for reliable HS-Gear5 link training, and the MCQ resource/runtime-config hooks required forufshcd_alloc_mcq()to succeed. Without either, the controller still works, but silently falls back to Gear5 link instability or legacy single-doorbell mode respectively.
Applied to 7.3/scsi-staging, thanks!
Signature · 1 line
Martin K. Petersen
Quoted 11 lines · sangram kumar yerra
096cd6b7adf2("scsi: ufs: ufs-pci: Add support for Intel Nova Lake") added the PCI ID for Intel's UFS 4.0 controller (8086:D335), reusing the existing Meteor Lake variant operations table. That table is missing two things this controller needs to work correctly at UFS 4.0
capability: aPA_INITIAL_ADAPTsequence for reliable HS-Gear5 link training, and the MCQ resource/runtime-config hooks required forufshcd_alloc_mcq()to succeed. Without either, the controller still works, but silently falls back to Gear5 link instability or legacy single-doorbell mode respectively.[...]
Applied to 7.3/scsi-fixes, thanks!
[1/2] scsi: ufs: ufs-pci: Add support for Intel UFS 4.0 HS-Gear5
https://git.kernel.org/mkp/scsi/c/ef675ea16845 [2/2] scsi: ufs: ufs-pci: Add MCQ support for Intel UFS 4.0 controllers https://git.kernel.org/mkp/scsi/c/c46cc9cee39b
Signature · 1 line
Martin K. Petersen