[Intel-xe] [PATCH 2/2] drm/xe: Sysfs entries to query vram fused min, max frequency
Gupta, Anshuman
anshuman.gupta at intel.com
Mon Aug 21 13:25:47 UTC 2023
> -----Original Message-----
> From: Intel-xe <intel-xe-bounces at lists.freedesktop.org> On Behalf Of
> Sujaritha Sundaresan
> Sent: Monday, August 21, 2023 5:37 PM
> To: intel-xe at lists.freedesktop.org
> Cc: Sundaresan, Sujaritha <sujaritha.sundaresan at intel.com>
> Subject: [Intel-xe] [PATCH 2/2] drm/xe: Sysfs entries to query vram fused
> min, max frequency
>
> Add sysfs entries to query fused min and max frequency of vram
>
> Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan at intel.com>
> ---
> drivers/gpu/drm/xe/xe_guc_pc.c | 55
> +++++++++++++++++++++++++++++++
> drivers/gpu/drm/xe/xe_pcode_api.h | 8 +++++
> 2 files changed, 63 insertions(+)
>
> diff --git a/drivers/gpu/drm/xe/xe_guc_pc.c
> b/drivers/gpu/drm/xe/xe_guc_pc.c index c03bb58e7049..d7e9609c74c0
> 100644
> --- a/drivers/gpu/drm/xe/xe_guc_pc.c
> +++ b/drivers/gpu/drm/xe/xe_guc_pc.c
> @@ -20,6 +20,7 @@
> #include "xe_map.h"
> #include "xe_mmio.h"
> #include "xe_pcode.h"
> +#include "xe_pcode_api.h"
>
> #define MCHBAR_MIRROR_BASE_SNB 0x140000
>
> @@ -568,6 +569,54 @@ static ssize_t freq_max_store(struct device *dev,
> struct device_attribute *attr, } static DEVICE_ATTR_RW(freq_max);
>
> +static ssize_t freq_vram_rp0_show(struct device *dev, struct
> device_attribute *attr,
> + char *buff)
> +{
> + struct xe_guc_pc *pc = dev_to_pc(dev);
> + struct xe_gt *gt = pc_to_gt(pc);
> + u32 val;
> + int err;
> +
> + err = xe_pcode_read_p(gt, XEHP_PCODE_FREQUENCY_CONFIG,
> + PCODE_MBOX_FC_SC_READ_FUSED_P0,
> + PCODE_MBOX_DOMAIN_HBM, &val);
> + if (err)
> + return err;
> +
> + /* data_out - Fused P0 for domain ID in units of 50 MHz */
> + val *= GT_FREQUENCY_MULTIPLIER;
> +
> + return sysfs_emit(buff, "%u\n", val);
> +}
> +static DEVICE_ATTR_RO(freq_vram_rp0);
> +
> +static ssize_t freq_vram_rpn_show(struct device *dev, struct
> device_attribute *attr,
> + char *buff)
> +{
> + struct xe_guc_pc *pc = dev_to_pc(dev);
> + struct xe_gt *gt = pc_to_gt(pc);
> + u32 val;
> + int err;
> +
> + err = xe_pcode_read_p(gt, XEHP_PCODE_FREQUENCY_CONFIG,
> + PCODE_MBOX_FC_SC_READ_FUSED_PN,
> + PCODE_MBOX_DOMAIN_HBM, &val);
> + if (err)
> + return err;
> +
> + /* data_out - Fused P0 for domain ID in units of 50 MHz */
> + val *= GT_FREQUENCY_MULTIPLIER;
> +
> + return sysfs_emit(buff, "%u\n", val);
> +}
> +static DEVICE_ATTR_RO(freq_vram_rpn);
> +
> +static const struct attribute *vram_freq_attrs[] = {
> + &dev_attr_freq_vram_rp0.attr,
> + &dev_attr_freq_vram_rpn.attr,
> + NULL
> +};
> +
> /**
> * xe_guc_pc_c_status - get the current GT C state
> * @pc: XE_GuC_PC instance
> @@ -921,6 +970,12 @@ int xe_guc_pc_init(struct xe_guc_pc *pc)
>
> pc_init_fused_rp_values(pc);
>
> + if (IS_DGFX(xe) && xe->info.platform != XE_DG2) {
Add this attribute only for PVC ?
> + err = sysfs_create_files(gt->sysfs, vram_freq_attrs);
vram attribute should be per tile instead of GT.
A tile can have multiple GT in theory, so associating vram attributes with GT is
theoretically wrong.
Cc: Matt Roper <matthew.d.roper at intel.com>
Thanks,
Anshuman Gupta.
> + if (err)
> + return err;
> + }
> +
> err = sysfs_create_files(gt->sysfs, pc_attrs);
> if (err)
> return err;
> diff --git a/drivers/gpu/drm/xe/xe_pcode_api.h
> b/drivers/gpu/drm/xe/xe_pcode_api.h
> index 837ff7c71280..846c99aba3bc 100644
> --- a/drivers/gpu/drm/xe/xe_pcode_api.h
> +++ b/drivers/gpu/drm/xe/xe_pcode_api.h
> @@ -25,6 +25,14 @@
> #define PCODE_DATA0 XE_REG(0x138128)
> #define PCODE_DATA1 XE_REG(0x13812C)
>
> +#define XEHP_PCODE_FREQUENCY_CONFIG 0x6e /* xehp, pvc
> */
> +/* XEHP_PCODE_FREQUENCY_CONFIG sub-commands (param1) */
> +#define PCODE_MBOX_FC_SC_READ_FUSED_P0 0x0
> +#define PCODE_MBOX_FC_SC_READ_FUSED_PN 0x1
> +/* PCODE_MBOX_DOMAIN_* - mailbox domain IDs */
> +/* XEHP_PCODE_FREQUENCY_CONFIG param2 */
> +#define PCODE_MBOX_DOMAIN_HBM 0x2
> +
> /* Min Freq QOS Table */
> #define PCODE_WRITE_MIN_FREQ_TABLE 0x8
> #define PCODE_READ_MIN_FREQ_TABLE 0x9
> --
> 2.25.1
More information about the Intel-xe
mailing list