From: Riana Tauro <riana.tauro@intel.com>
To: Soham Purkait <soham.purkait@intel.com>,
<intel-xe@lists.freedesktop.org>, <anshuman.gupta@intel.com>,
<badal.nilawar@intel.com>, <karthik.poosa@intel.com>
Cc: <lucas.demarchi@intel.com>, <ashutosh.dixit@intel.com>,
<jani.nikula@intel.com>
Subject: Re: [PATCH v3 1/2] drm/xe/xe_debugfs: Exposure of G-State and pcie link state residency counters through debugfs
Date: Thu, 5 Jun 2025 16:48:18 +0530 [thread overview]
Message-ID: <23fff9b4-a22b-4b8a-9d9c-45512ae96889@intel.com> (raw)
In-Reply-To: <20250603181205.1680363-2-soham.purkait@intel.com>
Hi Soham
On 6/3/2025 11:42 PM, Soham Purkait wrote:
> Add two debugfs node named dgfx_pkg_residencies and dgfx_link_state_residencys
> in order to obtain the G-State residency counter values for G2, G6, G8, G10 &
> ModS and the pcie link state residency counter values for L0, L1 & L1.2
> respectively.
>
> Signed-off-by: Soham Purkait <soham.purkait@intel.com>
> ---
> drivers/gpu/drm/xe/xe_debugfs.c | 99 +++++++++++++++++++++++++++++++++
> 1 file changed, 99 insertions(+)
>
> diff --git a/drivers/gpu/drm/xe/xe_debugfs.c b/drivers/gpu/drm/xe/xe_debugfs.c
> index d83cd6ed3fa8..c339f05ce42e 100644
> --- a/drivers/gpu/drm/xe/xe_debugfs.c
> +++ b/drivers/gpu/drm/xe/xe_debugfs.c
> @@ -11,16 +11,19 @@
>
> #include <drm/drm_debugfs.h>
>
> +#include "regs/xe_pmt.h"
> #include "xe_bo.h"
> #include "xe_device.h"
> #include "xe_force_wake.h"
> #include "xe_gt_debugfs.h"
> #include "xe_gt_printk.h"
> #include "xe_guc_ads.h"
> +#include "xe_mmio.h"
> #include "xe_pm.h"
> #include "xe_pxp_debugfs.h"
> #include "xe_sriov.h"
> #include "xe_step.h"
> +#include "xe_vsec.h"
>
> #ifdef CONFIG_DRM_XE_DEBUG
> #include "xe_bo_evict.h"
> @@ -185,6 +188,86 @@ static ssize_t wedged_mode_set(struct file *f, const char __user *ubuf,
> return size;
> }
>
> +static int read_residency_counter(struct xe_device *xe, struct xe_mmio *mmio,
> + u64 *dst, u32 offset, char *name)
> +{
> + int ret = xe_pmt_telem_read(to_pci_dev(xe->drm.dev),
> + xe_mmio_read32(mmio, PUNIT_TELEMETRY_GUID),
> + dst, offset, sizeof(u64));
What happens if this call fails? dst will be 0?
> + if (ret != sizeof(u64))
> + drm_warn(&xe->drm, "%s residency counter failed to read, ret %d\n", name, ret);
> +
> + return 0;
> +}
> +
> +static ssize_t dgfx_pkg_residencies_show(struct file *f, char __user *ubuf,
> + size_t size, loff_t *pos)
> +{
> + u64 g_states;
%s/g_states/residency
> + char buf[256];
> + int len = 0;
> + struct xe_device *xe;
> + struct xe_mmio *mmio;
> +
> + xe = file_inode(f)->i_private;
> + xe_pm_runtime_get(xe);
> + mmio = xe_root_tile_mmio(xe);
> +
> + g_states = 0;
> + read_residency_counter(xe, mmio, &g_states, BMG_G2_RESIDENCY_OFFSET, "G2");
> + len = scnprintf(buf, sizeof(buf), "Package G2: %llu\n", g_states);
> +
There is no check for BMG. Is it applicable only for BMG?
> + g_states = 0;
> + read_residency_counter(xe, mmio, &g_states, BMG_G6_RESIDENCY_OFFSET, "G6");
> + len += scnprintf(buf + len, sizeof(buf) - len, "Package G6: %llu\n", g_states);
> +
> + g_states = 0;
> + read_residency_counter(xe, mmio, &g_states, BMG_G8_RESIDENCY_OFFSET, "G8");
> + len += scnprintf(buf + len, sizeof(buf) - len, "Package G8: %llu\n", g_states);
> +
> + g_states = 0;
> + read_residency_counter(xe, mmio, &g_states, BMG_G10_RESIDENCY_OFFSET, "G10");
> + len += scnprintf(buf + len, sizeof(buf) - len, "Package G10: %llu\n", g_states);
> +
> + g_states = 0;
> + read_residency_counter(xe, mmio, &g_states, BMG_MODS_RESIDENCY_OFFSET, "ModS");
> + len += scnprintf(buf + len, sizeof(buf) - len, "Package ModS: %llu\n", g_states);
Repeated code here. Use a helper
> +
> + xe_pm_runtime_put(xe);
> + return simple_read_from_buffer(ubuf, size, pos, buf, len);
> +}
> +
> +static ssize_t dgfx_link_state_residencys_show(struct file *f, char __user *ubuf,
> + size_t size, loff_t *pos)
> +{
> + u64 link_res;
> + char buf[256];
> + int len = 0;
> + struct xe_device *xe;
> + struct xe_mmio *mmio;
> +
> + xe = file_inode(f)->i_private;
> + xe_pm_runtime_get(xe);
> + mmio = xe_root_tile_mmio(xe);
> +
> + link_res = 0;
> + read_residency_counter(xe, mmio, &link_res, PCIE_LINK_L0_RESIDENCY_COUNTER, "PCIE LINK L0");
> + len = scnprintf(buf, sizeof(buf), "PCIE LINK L0 RESIDENCY : %llu\n", link_res);
> +
> + link_res = 0;
> + read_residency_counter(xe, mmio, &link_res, PCIE_LINK_L1_RESIDENCY_COUNTER, "PCIE LINK L1");
> + len += scnprintf(buf + len, sizeof(buf) - len, "PCIE LINK L1 RESIDENCY : %llu\n", link_res);
> +
> + link_res = 0;
> + read_residency_counter(xe, mmio, &link_res,
> + PCIE_LINK_L1_2_RESIDENCY_COUNTER, "PCIE LINK L1.2");
> + len += scnprintf(buf + len, sizeof(buf) - len,
> + "PCIE LINK L1.2 RESIDENCY : %llu\n", link_res);
Use a helper
> +
> + xe_pm_runtime_put(xe);
> + return simple_read_from_buffer(ubuf, size, pos, buf, len);
> +}
> +
> static const struct file_operations wedged_mode_fops = {
> .owner = THIS_MODULE,
> .read = wedged_mode_show,
> @@ -226,6 +309,16 @@ static const struct file_operations atomic_svm_timeslice_ms_fops = {
> .write = atomic_svm_timeslice_ms_set,
> };
>
> +static const struct file_operations dgfx_pkg_residencies_fops = {
> + .owner = THIS_MODULE,
> + .read = dgfx_pkg_residencies_show,
> +};
> +
> +static const struct file_operations dgfx_link_state_residencys_fops = {
> + .owner = THIS_MODULE,
> + .read = dgfx_link_state_residencys_show,
> +};
> +
> void xe_debugfs_register(struct xe_device *xe)
> {
> struct ttm_device *bdev = &xe->ttm;
> @@ -249,6 +342,12 @@ void xe_debugfs_register(struct xe_device *xe)
> debugfs_create_file("atomic_svm_timeslice_ms", 0600, root, xe,
> &atomic_svm_timeslice_ms_fops);
>
> + debugfs_create_file("dgfx_pkg_residencies", 0444, root, xe,
> + &dgfx_pkg_residencies_fops);
> +
> + debugfs_create_file("dgfx_link_state_residencys", 0444, root, xe,
> + &dgfx_link_state_residencys_fops);
If these are for dgfx there needs to be dgfx check
Thanks
Riana
> +
> for (mem_type = XE_PL_VRAM0; mem_type <= XE_PL_VRAM1; ++mem_type) {
> man = ttm_manager_type(bdev, mem_type);
>
next prev parent reply other threads:[~2025-06-05 11:18 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-06-03 18:12 [PATCH v3 0/2] Add debugfs node to expose G-state and pcie link state residency Soham Purkait
2025-06-03 18:12 ` [PATCH v3 1/2] drm/xe/xe_debugfs: Exposure of G-State and pcie link state residency counters through debugfs Soham Purkait
2025-06-05 11:18 ` Riana Tauro [this message]
2025-06-05 13:23 ` Poosa, Karthik
2025-06-03 18:12 ` [PATCH v3 2/2] drm/xe/regs/xe_pmt: Macros for G-State and pcie link state residency offset Soham Purkait
2025-06-05 10:00 ` Poosa, Karthik
2025-06-03 18:22 ` ✓ CI.Patch_applied: success for Add debugfs node to expose G-state and pcie link state residency Patchwork
2025-06-03 18:22 ` ✗ CI.checkpatch: warning " Patchwork
2025-06-03 18:23 ` ✗ CI.KUnit: failure " Patchwork
2025-06-05 10:02 ` [PATCH v3 0/2] " Poosa, Karthik
2025-06-05 10:15 ` Poosa, Karthik
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=23fff9b4-a22b-4b8a-9d9c-45512ae96889@intel.com \
--to=riana.tauro@intel.com \
--cc=anshuman.gupta@intel.com \
--cc=ashutosh.dixit@intel.com \
--cc=badal.nilawar@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=jani.nikula@intel.com \
--cc=karthik.poosa@intel.com \
--cc=lucas.demarchi@intel.com \
--cc=soham.purkait@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox