[Intel-xe] [PATCH V2 2/3] drm/xe/ats-m: Expose uid for ATS-M via sysfs
Upadhyay, Tejas
tejas.upadhyay at intel.com
Wed Nov 29 05:09:12 UTC 2023
> -----Original Message-----
> From: De Marchi, Lucas <lucas.demarchi at intel.com>
> Sent: Wednesday, November 29, 2023 12:12 AM
> To: Upadhyay, Tejas <tejas.upadhyay at intel.com>
> Cc: intel-xe at lists.freedesktop.org
> Subject: Re: [Intel-xe] [PATCH V2 2/3] drm/xe/ats-m: Expose uid for ATS-M via
> sysfs
>
> On Fri, Nov 24, 2023 at 07:43:10PM +0530, Tejas Upadhyay wrote:
> >On ATS-M CSC firmware calculated unique device id(uid) can be read from
> >defined registers. Only ATS-M has this way of reading uid. Also ATS-M
> >customer require this solution to locate and name their devices
> >effectively.
>
> s/Also ATS-M.*//
>
> As this justification doesn't really belong to a commit message.
>
> *This* solution can't be a requirement. *A* solution, with this being the
> proposed one, maybe. Imagine the mess if each customer with each platform
> requires a different solution and each solution only works with one of them.
>
> >
> >Consumer of csc uid is level0 sysman and end users using sysman.
>
> so you mean that only ATS-M works with level0 sysman? That seems a very
> poor abstraction to be done in the kernel. If we need a uid, let's give a uid
> that can support different backends so we don't confine the support to 1
> platform for no reason.
No, By this I was meaning, ATS-M consumer seems to have requirement of reading UID, which they are going to get through sysman.
Also, In future when there is requirement on other platforms to read UID, then xe_read_csc_uid() API can be made generic like xe_read_ uid () to support all types of methods to read UID.
Infact I will change it to make it generic as part of next patchset with only abstraction being UID read method in xe_read_csc_uid().
Thanks,
Tejas
>
> Lucas De Marchi
>
> >
> >This patch exports uid to the userspace via sysfs.
> >For example, uid can be read by:
> >cat /sys/class/drm/cardX/device/csc_uid
> >
> >V2(Michal):
> > - Make separate patch for csc uid capability
> > - Use %#llx for 0x
> > - Dump error value when sysfs creation fails
> > - Have consistency in csc uid name
> >
> >Signed-off-by: Tejas Upadhyay <tejas.upadhyay at intel.com>
> >---
> > drivers/gpu/drm/xe/regs/xe_regs.h | 3 +++
> > drivers/gpu/drm/xe/xe_device.c | 11 +++++++++++
> > drivers/gpu/drm/xe/xe_device_sysfs.c | 21 +++++++++++++++++++++
> >drivers/gpu/drm/xe/xe_device_types.h | 2 ++
> > 4 files changed, 37 insertions(+)
> >
> >diff --git a/drivers/gpu/drm/xe/regs/xe_regs.h
> >b/drivers/gpu/drm/xe/regs/xe_regs.h
> >index ec9372aa739f..3599c7b38b61 100644
> >--- a/drivers/gpu/drm/xe/regs/xe_regs.h
> >+++ b/drivers/gpu/drm/xe/regs/xe_regs.h
> >@@ -71,6 +71,9 @@
> > #define XEHP_CLOCK_GATE_DIS XE_REG(0x101014)
> > #define SGSI_SIDECLK_DIS REG_BIT(17)
> >
> >+#define DEVUID_LWORD XE_REG(0x102008)
> >+#define DEVUID_HWORD XE_REG(0x10200C)
> >+
> > #define GGC XE_REG(0x108040)
> > #define GMS_MASK REG_GENMASK(15, 8)
> > #define GGMS_MASK REG_GENMASK(7, 6)
> >diff --git a/drivers/gpu/drm/xe/xe_device.c
> >b/drivers/gpu/drm/xe/xe_device.c index 8be765adf702..6c8a9f9d69c9
> >100644
> >--- a/drivers/gpu/drm/xe/xe_device.c
> >+++ b/drivers/gpu/drm/xe/xe_device.c
> >@@ -342,6 +342,15 @@ static void xe_device_sanitize(struct drm_device
> *drm, void *arg)
> > xe_gt_sanitize(gt);
> > }
> >
> >+static void xe_read_csc_uid(struct xe_device *xe) {
> >+ if (xe->info.has_csc_uid) {
> >+ struct xe_gt *mmio = xe_root_mmio_gt(xe);
> >+
> >+ xe->info.csc_uid = xe_mmio_read64_2x32(mmio,
> DEVUID_LWORD);
> >+ }
> >+}
> >+
> > int xe_device_probe(struct xe_device *xe) {
> > struct xe_tile *tile;
> >@@ -366,6 +375,8 @@ int xe_device_probe(struct xe_device *xe)
> > if (err)
> > return err;
> >
> >+ xe_read_csc_uid(xe);
> >+
> > err = drmm_add_action_or_reset(&xe->drm, xe_driver_flr_fini, xe);
> > if (err)
> > return err;
> >diff --git a/drivers/gpu/drm/xe/xe_device_sysfs.c
> >b/drivers/gpu/drm/xe/xe_device_sysfs.c
> >index 99113a5a2b84..4fbf0275d3bf 100644
> >--- a/drivers/gpu/drm/xe/xe_device_sysfs.c
> >+++ b/drivers/gpu/drm/xe/xe_device_sysfs.c
> >@@ -65,11 +65,24 @@ vram_d3cold_threshold_store(struct device *dev,
> >struct device_attribute *attr,
> >
> > static DEVICE_ATTR_RW(vram_d3cold_threshold);
> >
> >+static ssize_t csc_uid_show(struct device *dev,
> >+ struct device_attribute *attr,
> >+ char *buf)
> >+{
> >+ struct pci_dev *pdev = to_pci_dev(dev);
> >+ struct xe_device *xe = pdev_to_xe_device(pdev);
> >+
> >+ return sysfs_emit(buf, "%#llx\n", xe->info.csc_uid); }
> >+
> >+static DEVICE_ATTR_RO(csc_uid);
> >+
> > static void xe_device_sysfs_fini(struct drm_device *drm, void *arg) {
> > struct xe_device *xe = arg;
> >
> > sysfs_remove_file(&xe->drm.dev->kobj,
> > &dev_attr_vram_d3cold_threshold.attr);
> >+ sysfs_remove_file(&xe->drm.dev->kobj, &dev_attr_csc_uid.attr);
> > }
> >
> > void xe_device_sysfs_init(struct xe_device *xe) @@ -83,6 +96,14 @@
> >void xe_device_sysfs_init(struct xe_device *xe)
> > return;
> > }
> >
> >+ if (xe->info.has_csc_uid) {
> >+ ret = sysfs_create_file(&dev->kobj, &dev_attr_csc_uid.attr);
> >+ if (ret) {
> >+ drm_warn(&xe->drm, "UID sysfs setup failed
> err=%pe\n", ERR_PTR(ret));
> >+ return;
> >+ }
> >+ }
> >+
> > ret = drmm_add_action_or_reset(&xe->drm, xe_device_sysfs_fini,
> xe);
> > if (ret)
> > drm_warn(&xe->drm, "Failed to add sysfs fini drm action\n");
> diff
> >--git a/drivers/gpu/drm/xe/xe_device_types.h
> >b/drivers/gpu/drm/xe/xe_device_types.h
> >index a28257b26355..eb3897309dac 100644
> >--- a/drivers/gpu/drm/xe/xe_device_types.h
> >+++ b/drivers/gpu/drm/xe/xe_device_types.h
> >@@ -238,6 +238,8 @@ struct xe_device {
> > u8 vm_max_level;
> > /** @va_bits: Maximum bits of a virtual address */
> > u8 va_bits;
> >+ /** @csc_uid: device uid calculated by CSC FW, used for
> generating uuid */
> >+ u64 csc_uid;
> >
> > /** @is_dgfx: is discrete device */
> > u8 is_dgfx:1;
> >--
> >2.25.1
> >
More information about the Intel-xe
mailing list