Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Matthew Brost <matthew.brost@intel.com>
To: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>
Cc: <intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH v6 23/24] drm/xe: Add modparam to enable / disable ULLS on migrate queue
Date: Wed, 9 Sep 2026 11:11:59 -0700	[thread overview]
Message-ID: <aqGhb9k93U7PoJ1L@gsse-cloud1.jf.intel.com> (raw)
In-Reply-To: <ede219e02fab3c78c66edfb9cabd0e029de318e6.camel@linux.intel.com>

On Wed, Sep 09, 2026 at 10:03:32AM +0200, Thomas Hellström wrote:
> On Fri, 2026-09-04 at 14:16 -0700, Matthew Brost wrote:
> > Having modparam to enable / disable ULLS on migrate queue will help
> > with
> > quick experiments.
> > 
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > Link:
> > https://patch.msgid.link/20260228013501.106680-26-matthew.brost@intel.com
> > Signed-off-by: Maarten Lankhorst <dev@lankhorst.se>
> > ---
> >  drivers/gpu/drm/xe/xe_debugfs.c      | 1 +
> >  drivers/gpu/drm/xe/xe_defaults.h     | 1 +
> >  drivers/gpu/drm/xe/xe_device.c       | 1 +
> >  drivers/gpu/drm/xe/xe_device_types.h | 5 +++++
> >  drivers/gpu/drm/xe/xe_migrate.c      | 2 +-
> >  drivers/gpu/drm/xe/xe_module.c       | 4 ++++
> >  drivers/gpu/drm/xe/xe_module.h       | 1 +
> >  7 files changed, 14 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_debugfs.c
> > b/drivers/gpu/drm/xe/xe_debugfs.c
> > index 80f62634fae5..f3f37d116d62 100644
> > --- a/drivers/gpu/drm/xe/xe_debugfs.c
> > +++ b/drivers/gpu/drm/xe/xe_debugfs.c
> > @@ -222,6 +222,7 @@ static int info(struct seq_file *m, void *data)
> >  	drm_printf(&p, "revid %d\n", xe->info.revid);
> >  	drm_printf(&p, "tile_count %d\n", xe->info.tile_count);
> >  	drm_printf(&p, "vm_max_level %d\n", xe->info.vm_max_level);
> > +	drm_printf(&p, "ulls_enable %s\n", str_yes_no(xe-
> > >info.ulls_enable));
> >  	drm_printf(&p, "has_flat_ccs %s\n", str_yes_no(xe-
> > >info.has_flat_ccs));
> >  	drm_printf(&p, "has_usm %s\n", str_yes_no(xe-
> > >info.has_usm));
> >  	drm_printf(&p, "skip_guc_pc %s\n", str_yes_no(xe-
> > >info.skip_guc_pc));
> > diff --git a/drivers/gpu/drm/xe/xe_defaults.h
> > b/drivers/gpu/drm/xe/xe_defaults.h
> > index 0884224ef7c7..9c7e89a169c5 100644
> > --- a/drivers/gpu/drm/xe/xe_defaults.h
> > +++ b/drivers/gpu/drm/xe/xe_defaults.h
> > @@ -14,6 +14,7 @@
> >  #endif
> >  
> >  #define
> > XE_DEFAULT_PROBE_DISPLAY		IS_ENABLED(CONFIG_DRM_XE_DISPLAY)
> > +#define XE_DEFAULT_ULLS_ENABLE			true
> >  #define XE_DEFAULT_VRAM_BAR_SIZE		0
> >  #define
> > XE_DEFAULT_FORCE_PROBE			CONFIG_DRM_XE_FORCE_PROBE
> >  #define XE_DEFAULT_MAX_VFS			~0
> > diff --git a/drivers/gpu/drm/xe/xe_device.c
> > b/drivers/gpu/drm/xe/xe_device.c
> > index ade971de1652..9b81759af28a 100644
> > --- a/drivers/gpu/drm/xe/xe_device.c
> > +++ b/drivers/gpu/drm/xe/xe_device.c
> > @@ -520,6 +520,7 @@ static void xe_device_parse_modparam(struct
> > xe_device *xe)
> >  	xe->atomic_svm_timeslice_ms = 5;
> >  	xe->min_run_period_lr_ms = 5;
> >  	xe->info.num_pf_work = xe_modparam.num_pf_work;
> > +	xe->info.ulls_enable = xe_modparam.ulls_enable;
> 
> migrate_ulls_enable?
> 
> >  	if (xe->info.num_pf_work < 1)
> >  		xe->info.num_pf_work = 1;
> >  	else if (xe->info.num_pf_work > XE_PAGEFAULT_WORK_MAX)
> > diff --git a/drivers/gpu/drm/xe/xe_device_types.h
> > b/drivers/gpu/drm/xe/xe_device_types.h
> > index 997ac82fd571..fd4560378f5b 100644
> > --- a/drivers/gpu/drm/xe/xe_device_types.h
> > +++ b/drivers/gpu/drm/xe/xe_device_types.h
> > @@ -245,6 +245,11 @@ struct xe_device {
> >  		u8 skip_pcode:1;
> >  		/** @info.needs_shared_vf_gt_wq: needs shared GT WQ
> > on VF */
> >  		u8 needs_shared_vf_gt_wq:1;
> > +		/**
> > +		 * @info.ulls_enable: Enable ULLS on migration queue
> > in LR VM
> > +		 * open
> > +		 */
> > +		u8 ulls_enable:1;
> >  	} info;
> >  
> >  	/** @wa_active: keep track of active workarounds */
> > diff --git a/drivers/gpu/drm/xe/xe_migrate.c
> > b/drivers/gpu/drm/xe/xe_migrate.c
> > index 3e59aeeca614..3bc78f761f23 100644
> > --- a/drivers/gpu/drm/xe/xe_migrate.c
> > +++ b/drivers/gpu/drm/xe/xe_migrate.c
> > @@ -523,7 +523,7 @@ void xe_migrate_ulls_enter(struct xe_migrate *m)
> >  
> >  	xe_assert(xe, xe->info.has_usm);
> >  
> > -	if (!IS_DGFX(xe))
> > +	if (!IS_DGFX(xe) || !xe->info.ulls_enable)
> >  		return;
> >  
> >  job_alloc:
> > diff --git a/drivers/gpu/drm/xe/xe_module.c
> > b/drivers/gpu/drm/xe/xe_module.c
> > index 4bc28dfc1992..a8b546550ead 100644
> > --- a/drivers/gpu/drm/xe/xe_module.c
> > +++ b/drivers/gpu/drm/xe/xe_module.c
> > @@ -23,6 +23,7 @@
> >  
> >  struct xe_modparam xe_modparam = {
> >  	.probe_display =	XE_DEFAULT_PROBE_DISPLAY,
> > +	.ulls_enable =		XE_DEFAULT_ULLS_ENABLE,
> >  	.guc_log_level =	XE_DEFAULT_GUC_LOG_LEVEL,
> >  	.force_probe =		XE_DEFAULT_FORCE_PROBE,
> >  #ifdef CONFIG_PCI_IOV
> > @@ -44,6 +45,9 @@ MODULE_PARM_DESC(probe_display, "Probe display HW,
> > otherwise it's left untouched
> >  		 "[default=" __stringify(XE_DEFAULT_PROBE_DISPLAY)
> > "])");
> >  #endif
> >  
> > +module_param_named(ulls_enable, xe_modparam.ulls_enable, bool,
> > 0444);
> > +MODULE_PARM_DESC(ulls_enable, "Enable ULLS on migration queue if LR
> > VM open (default: true)");
> > +
> >  module_param_named(vram_bar_size, xe_modparam.force_vram_bar_size,
> > int, 0600);
> >  MODULE_PARM_DESC(vram_bar_size, "Set the vram bar size in MiB
> > (<0=disable-resize, 0=max-needed-size, >0=force-size "
> >  		 "[default=" __stringify(XE_DEFAULT_VRAM_BAR_SIZE)
> > "])");
> > diff --git a/drivers/gpu/drm/xe/xe_module.h
> > b/drivers/gpu/drm/xe/xe_module.h
> > index 6272d9e41207..5d88448c298a 100644
> > --- a/drivers/gpu/drm/xe/xe_module.h
> > +++ b/drivers/gpu/drm/xe/xe_module.h
> > @@ -13,6 +13,7 @@ struct work_struct;
> >  /* Module modprobe variables */
> >  struct xe_modparam {
> >  	bool probe_display;
> > +	bool ulls_enable;
> 
> Shouldn't this be per device rather than per driver? So sysfs knob,
> perhaps with a modparam default?
> 

I didn't really want to add a userspace ABI, such as a sysfs interface,
at this point, so I went with a module parameter instead. It also seems
a bit odd to enable this on some devices but not others.

The main reason for having this at all is to allow quick performance
comparisons and, of course, to give us an easy way to disable it and
retest if we encounter workloads where kernel jobs appear to hang.

I'm open to changing this as the series receives more review and we get
closer to merging.

Matt

> Thanks,
> Thomas
> 
> 
> 
> >  	int force_vram_bar_size;
> >  	int guc_log_level;
> >  	char *guc_firmware_path;

  reply	other threads:[~2026-09-09 18:12 UTC|newest]

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-04 21:15 [PATCH v6 00/24] CPU binds and ULLS on migration queue Matthew Brost
2026-09-04 21:15 ` [PATCH v6 01/24] drm/xe: Drop struct xe_migrate_pt_update argument from populate/clear vfuns Matthew Brost
2026-09-04 21:15 ` [PATCH v6 02/24] drm/xe: Add xe_migrate_update_pgtables_cpu_execute helper Matthew Brost
2026-09-04 21:28   ` sashiko-bot
2026-09-04 21:15 ` [PATCH v6 03/24] drm/xe: Decouple exec queue idle check from LRC Matthew Brost
2026-09-04 21:15 ` [PATCH v6 04/24] drm/xe: Add job count to GuC exec queue snapshot Matthew Brost
2026-09-04 21:23   ` sashiko-bot
2026-09-04 21:15 ` [PATCH v6 05/24] drm/xe: Update xe_bo_put_deferred arguments to include writeback flag Matthew Brost
2026-09-04 21:15 ` [PATCH v6 06/24] drm/xe: Add XE_BO_FLAG_PUT_VM_ASYNC Matthew Brost
2026-09-04 21:33   ` sashiko-bot
2026-09-11 13:10   ` Francois Dugast
2026-09-11 19:54     ` Matthew Brost
2026-09-12  0:27       ` Matthew Brost
2026-09-04 21:15 ` [PATCH v6 07/24] drm/xe: Update scheduler job layer to support PT jobs Matthew Brost
2026-09-04 21:37   ` sashiko-bot
2026-09-11 15:24   ` Francois Dugast
2026-09-11 19:25     ` Matthew Brost
2026-09-04 21:15 ` [PATCH v6 08/24] drm/xe: Add helpers to access PT ops Matthew Brost
2026-09-04 21:15 ` [PATCH v6 09/24] drm/xe: Add struct xe_pt_job_ops Matthew Brost
2026-09-04 21:40   ` sashiko-bot
2026-09-04 21:15 ` [PATCH v6 10/24] drm/xe: Update GuC submission backend to run PT jobs Matthew Brost
2026-09-04 21:39   ` sashiko-bot
2026-09-04 21:16 ` [PATCH v6 11/24] drm/xe: Store level in struct xe_vm_pgtable_update Matthew Brost
2026-09-04 21:16 ` [PATCH v6 12/24] drm/xe: Don't use migrate exec queue for page fault binds Matthew Brost
2026-09-04 21:16 ` [PATCH v6 13/24] drm/xe: Enable CPU binds for jobs Matthew Brost
2026-09-04 21:44   ` sashiko-bot
2026-09-04 21:16 ` [PATCH v6 14/24] drm/xe: Remove unused arguments from xe_migrate_pt_update_ops Matthew Brost
2026-09-04 21:16 ` [PATCH v6 15/24] drm/xe: Make bind queues operate cross-tile Matthew Brost
2026-09-04 21:16 ` [PATCH v6 16/24] drm/xe: Add CPU bind layer Matthew Brost
2026-09-04 21:50   ` sashiko-bot
2026-09-04 21:16 ` [PATCH v6 17/24] drm/xe: Add device flag to enable PT mirroring across tiles Matthew Brost
2026-09-04 21:40   ` sashiko-bot
2026-09-04 21:16 ` [PATCH v6 18/24] drm/xe: Add ULLS support to LRC Matthew Brost
2026-09-04 21:16 ` [PATCH v6 19/24] drm/xe: Add ULLS migration job support to migration layer Matthew Brost
2026-09-04 21:40   ` sashiko-bot
2026-09-04 21:16 ` [PATCH v6 20/24] drm/xe: Add ULLS migration job support to ring ops Matthew Brost
2026-09-04 21:16 ` [PATCH v6 21/24] drm/xe: Add ULLS migration job support to GuC submission Matthew Brost
2026-09-04 21:16 ` [PATCH v6 22/24] drm/xe: Enter ULLS for migration jobs upon page fault or SVM prefetch Matthew Brost
2026-09-04 21:16 ` [PATCH v6 23/24] drm/xe: Add modparam to enable / disable ULLS on migrate queue Matthew Brost
2026-09-09  8:03   ` Thomas Hellström
2026-09-09 18:11     ` Matthew Brost [this message]
2026-09-04 21:16 ` [PATCH v6 24/24] drm/xe: Document ULLS for migration jobs Matthew Brost
2026-09-09  9:01   ` Thomas Hellström
2026-09-09 17:53     ` Matthew Brost
2026-09-04 21:24 ` ✗ CI.checkpatch: warning for CPU binds and ULLS on migration queue (rev8) Patchwork
2026-09-04 21:26 ` ✓ CI.KUnit: success " Patchwork
2026-09-04 22:16 ` ✓ Xe.CI.BAT: " Patchwork
2026-09-05  3:34 ` ✗ Xe.CI.FULL: failure " Patchwork

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=aqGhb9k93U7PoJ1L@gsse-cloud1.jf.intel.com \
    --to=matthew.brost@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=thomas.hellstrom@linux.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