mbox series

[v2,00/17] Add Multi Circular Queue Support

Message ID cover.1665017636.git.quic_asutoshd@quicinc.com
Headers show
Series Add Multi Circular Queue Support | expand

Message

Asutosh Das Oct. 6, 2022, 1:05 a.m. UTC
UFS Multi-Circular Queue (MCQ) has been added in UFSHCI v4.0 to improve storage performance.
The implementation uses the shared tagging mechanism so that tags are shared
among the hardware queues. The number of hardware queues is configurable.
This series doesn't include the ESI implementation for completion handling.

This implementation has been verified by booting on an emulation platform.
During testing, all low power modes were disabled and it was in HS-G1 mode.

Please take a look and let us know your thoughts.

v1 -> v2:
- Added a non MCQ related change to use a function to extrace ufs extended
feature
- Addressed Mani's comments
- Addressed Bart's comments

v1:
- Split the changes
- Addressed Bart's comments
- Addressed Bean's comments

* RFC versions:
v2 -> v3:
- Split the changes based on functionality
- Addressed queue configuration issues
- Faster SQE tail pointer increments
- Addressed comments from Bart and Manivannan

v1 -> v2:
- Enabled host_tagset
- Added queue num configuration support
- Added one more vops to allow vendor provide the wanted MAC
- Determine nutrs and can_queue by considering both MAC, bqueuedepth and EXT_IID support
- Postponed MCQ initialization and scsi_add_host() to async probe
- Used (EXT_IID, Task Tag) tuple to support up to 4096 tasks (theoretically)

Asutosh Das (17):
  ufs: core: Probe for ext_iid support
  ufs: core: Optimize duplicate code to read extended feature
  ufs: core: Introduce Multi-circular queue capability
  ufs: core: Defer adding host to scsi if mcq is supported
  ufs: core: mcq: Introduce Multi Circular Queue
  ufs: core: mcq: Configure resource regions
  ufs: core: mcq: Calculate queue depth
  ufs: core: mcq: Allocate memory for mcq mode
  ufs: core: mcq: Configure operation and runtime interface
  ufs: core: mcq: Use shared tags for MCQ mode
  ufs: core: Prepare ufshcd_send_command for mcq
  ufs: core: mcq: Find hardware queue to queue request
  ufs: core: Prepare for completion in mcq
  ufs: mcq: Add completion support of a cqe
  ufs: core: mcq: Add completion support in poll
  ufs: core: mcq: Enable Multi Circular Queue
  ufs: qcom-host: Enable multi circular queue capability

 drivers/ufs/core/Makefile      |   2 +-
 drivers/ufs/core/ufs-mcq.c     | 495 +++++++++++++++++++++++++++++++++++++++++
 drivers/ufs/core/ufshcd-priv.h |  84 ++++++-
 drivers/ufs/core/ufshcd.c      | 345 ++++++++++++++++++++++------
 drivers/ufs/host/ufs-qcom.c    |  49 ++++
 drivers/ufs/host/ufs-qcom.h    |   4 +
 include/ufs/ufs.h              |   6 +
 include/ufs/ufshcd.h           | 136 +++++++++++
 include/ufs/ufshci.h           |  63 ++++++
 9 files changed, 1116 insertions(+), 68 deletions(-)
 create mode 100644 drivers/ufs/core/ufs-mcq.c

Comments

Daejun Park Oct. 7, 2022, 3:44 a.m. UTC | #1
Hi Asutosh Das,

>Define the mcq resources and add support to ioremap
>the resource regions.
>
>Co-developed-by: Can Guo <quic_cang@quicinc.com>
>Signed-off-by: Can Guo <quic_cang@quicinc.com>
>Signed-off-by: Asutosh Das <quic_asutoshd@quicinc.com>
>---
> drivers/ufs/core/ufs-mcq.c | 99 ++++++++++++++++++++++++++++++++++++++++++++++
> include/ufs/ufshcd.h       | 28 +++++++++++++
> 2 files changed, 127 insertions(+)
>
>diff --git a/drivers/ufs/core/ufs-mcq.c b/drivers/ufs/core/ufs-mcq.c
>index 659398d..7d0a37a 100644
>--- a/drivers/ufs/core/ufs-mcq.c
>+++ b/drivers/ufs/core/ufs-mcq.c
>@@ -18,6 +18,11 @@
> #define UFS_MCQ_NUM_DEV_CMD_QUEUES 1
> #define UFS_MCQ_MIN_POLL_QUEUES 0

>+#define MCQ_QCFGPTR_MASK        GENMASK(7, 0)
>+#define MCQ_QCFGPTR_UNIT        0x200
>+#define MCQ_SQATTR_OFFSET(c) \
>+        ((((c) >> 16) & MCQ_QCFGPTR_MASK) * MCQ_QCFGPTR_UNIT)
>+#define MCQ_QCFG_SIZE        0x40

> static int rw_queue_count_set(const char *val, const struct kernel_param *kp)
> {
>@@ -67,6 +72,97 @@ module_param_cb(poll_queues, &poll_queue_count_ops, &poll_queues, 0644);
> MODULE_PARM_DESC(poll_queues,
>                  "Number of poll queues used for r/w. Default value is 1");

>+/* Resources */
>+static const struct ufshcd_res_info ufs_res_info[RES_MAX] = {
>+        {.name = "ufs_mem",},
>+        {.name = "mcq",},
>+        /* Submission Queue DAO */
>+        {.name = "mcq_sqd",},
>+        /* Submission Queue Interrupt Status */
>+        {.name = "mcq_sqis",},
>+        /* Completion Queue DAO */
>+        {.name = "mcq_cqd",},
>+        /* Completion Queue Interrupt Status */
>+        {.name = "mcq_cqis",},
>+        /* MCQ vendor specific */
>+        {.name = "mcq_vs",},
>+};
>+
>+static int ufshcd_mcq_config_resource(struct ufs_hba *hba)
>+{
>+        struct platform_device *pdev = to_platform_device(hba->dev);
>+        struct ufshcd_res_info *res;
>+        struct resource *res_mem, *res_mcq;
>+        int i, ret = 0;
>+
>+        memcpy(hba->res, ufs_res_info, sizeof(ufs_res_info));
>+
>+        for (i = 0; i < RES_MAX; i++) {
>+                res = &hba->res[i];
>+                res->resource = platform_get_resource_byname(pdev,
>+                                                             IORESOURCE_MEM,
>+                                                             res->name);
>+                if (!res->resource) {
>+                        dev_info(hba->dev, "Resource %s not provided\n", res->name);
>+                        if (i == RES_UFS)
>+                                return -ENOMEM;
>+                        continue;
>+                } else if (i == RES_UFS) {
>+                        res_mem = res->resource;
>+                        res->base = hba->mmio_base;
>+                        continue;
>+                }
>+
>+                res->base = devm_ioremap_resource(hba->dev, res->resource);
>+                if (IS_ERR(res->base)) {
>+                        dev_err(hba->dev, "Failed to map res %s, err=%d\n",
>+                                         res->name, (int)PTR_ERR(res->base));
>+                        res->base = NULL;
>+                        ret = PTR_ERR(res->base);
>+                        return ret;
>+                }
>+        }
>+
>+        /* MCQ resource provided in DT */
>+        res = &hba->res[RES_MCQ];
>+        /* Bail if NCQ resource is provided */
Maybe MCQ?

>+        if (res->base)
>+                goto out;
>+
>+        /* Manually allocate MCQ resource from ufs_mem */
>+        res_mcq = res->resource;
Why we assign the value to res_mcq?

>+        res_mcq = devm_kzalloc(hba->dev, sizeof(*res_mcq), GFP_KERNEL);
>+        if (!res_mcq) {
>+                dev_err(hba->dev, "Failed to allocate MCQ resource\n");
>+                return ret;
>+        }
>+
>+        res_mcq->start = res_mem->start +
>+                         MCQ_SQATTR_OFFSET(hba->mcq_capabilities);
>+        res_mcq->end = res_mcq->start + hba->nr_hw_queues * MCQ_QCFG_SIZE - 1;
>+        res_mcq->flags = res_mem->flags;
>+        res_mcq->name = "mcq";
>+
>+        ret = insert_resource(&iomem_resource, res_mcq);
>+        if (ret) {
>+                dev_err(hba->dev, "Failed to insert MCQ resource, err=%d\n", ret);
Should we free the res_mcq?

>+                return ret;
>+        }
>+
>+        res->base = devm_ioremap_resource(hba->dev, res_mcq);
>+        if (IS_ERR(res->base)) {
>+                dev_err(hba->dev, "MCQ registers mapping failed, err=%d\n",
>+                        (int)PTR_ERR(res->base));
Should we call remove_resource and free the res_mcq?

Thanks,
Daejun
Asutosh Das Oct. 17, 2022, 9:03 p.m. UTC | #2
On Fri, Oct 07 2022 at 20:44 -0700, Daejun Park wrote:
>Hi Asutosh Das,
>
[...]
>>+        res = &hba->res[RES_MCQ];
>>+        /* Bail if NCQ resource is provided */
>Maybe MCQ?
Thanks, will fix it.

>
>>+        if (res->base)
>>+                goto out;
>>+
>>+        /* Manually allocate MCQ resource from ufs_mem */
>>+        res_mcq = res->resource;
>Why we assign the value to res_mcq?
>
Hmm, good point, will fix it.

>>+        res_mcq = devm_kzalloc(hba->dev, sizeof(*res_mcq), GFP_KERNEL);
>>+        if (!res_mcq) {
>>+                dev_err(hba->dev, "Failed to allocate MCQ resource\n");
>>+                return ret;
>>+        }
>>+
>>+        res_mcq->start = res_mem->start +
>>+                         MCQ_SQATTR_OFFSET(hba->mcq_capabilities);
>>+        res_mcq->end = res_mcq->start + hba->nr_hw_queues * MCQ_QCFG_SIZE - 1;
>>+        res_mcq->flags = res_mem->flags;
>>+        res_mcq->name = "mcq";
>>+
>>+        ret = insert_resource(&iomem_resource, res_mcq);
>>+        if (ret) {
>>+                dev_err(hba->dev, "Failed to insert MCQ resource, err=%d\n", ret);
>Should we free the res_mcq?
>
yes, will add it. Thanks.

>>+                return ret;
>>+        }
>>+
>>+        res->base = devm_ioremap_resource(hba->dev, res_mcq);
>>+        if (IS_ERR(res->base)) {
>>+                dev_err(hba->dev, "MCQ registers mapping failed, err=%d\n",
>>+                        (int)PTR_ERR(res->base));
>Should we call remove_resource and free the res_mcq?
>
I'll add the cleanup here.
>Thanks,
>Daejun