From: "Kumar G, Naresh" <naresh.kumar.g@intel.com>
To: "K V P, Satyanarayana" <satyanarayana.k.v.p@intel.com>,
<intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH v1 2/2] drm/xe: Mutual exclusivity between CCS-mode and PF
Date: Wed, 3 Dec 2025 18:13:26 +0530 [thread overview]
Message-ID: <b6500ae0-715a-4ee6-9cf8-e5ada3a09dbb@intel.com> (raw)
In-Reply-To: <39541b74-2fe5-47a7-bb97-76e3ff4799ba@intel.com>
Hi Satya,
On 03-12-2025 14:22, K V P, Satyanarayana wrote:
>
> On 28-Nov-25 10:48 PM, Nareshkumar Gollakoti wrote:
>> Due to SLA agreement between PF and VFs,the alternate CCS-mode
>> cannot be changed when VFs are already enabled.
>> Similarly, enabling VFs is not permitted when the alternate
>> CCS-mode is active.
>>
>> Signed-off-by: Nareshkumar Gollakoti<naresh.kumar.g@intel.com>
>> ---
>> drivers/gpu/drm/xe/xe_gt_ccs_mode.c | 59 ++++++++++++++++++++++-------
>> drivers/gpu/drm/xe/xe_gt_ccs_mode.h | 12 ++++++
>> 2 files changed, 57 insertions(+), 14 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_gt_ccs_mode.c b/drivers/gpu/drm/xe/xe_gt_ccs_mode.c
>> index e146e00b0ca2..6652c468e1be 100644
>> --- a/drivers/gpu/drm/xe/xe_gt_ccs_mode.c
>> +++ b/drivers/gpu/drm/xe/xe_gt_ccs_mode.c
>> @@ -13,6 +13,7 @@
>> #include "xe_gt_sysfs.h"
>> #include "xe_mmio.h"
>> #include "xe_sriov.h"
>> +#include "xe_sriov_pf.h"
>>
>> static void __xe_gt_apply_ccs_mode(struct xe_gt *gt, u32 num_engines)
>> {
>> @@ -108,6 +109,36 @@ ccs_mode_show(struct device *kdev,
>> return sysfs_emit(buf, "%u\n", gt->ccs_mode);
>> }
>>
>> +static int gt_prepare_ccs_mode_enabling(struct xe_gt *gt)
>> +{
>> + struct xe_device *xe = gt_to_xe(gt);
>> +
>> + if (!IS_SRIOV_PF(xe))
>> + return 0;
>
> The ccs mode will not be enabled for native mode (non-virtualized mode)
> with this. It should be something like
>
> if (IS_SRIOV_VF(xe))
>
> return 0;
>
> Same comment for other places as well.
As no CCS-mode sysfs entry for SRIOV VF Mode[1], no need check for VF mode.
[1]: https://patchwork.freedesktop.org/patch/691098/?series=158222&rev=1
In SRIOV PF mode, use xe_sriov_pf_lockdown to block CCS-mode. If VFs are
enabled. otherwise, CCS-mode can be allowed.
If not in PF mode, it is considered as Non-SRIOV mode and the function
returns 0. Therefore, CCS mode is enabled without any restrictions.
>
>> +
>> + /*
>> + * We can't change CCS-mode when VFs are already enabled
>> + * and we must prevent enabling VFs when alternate
>> + * CCS-mode is active
>> + */
>> + if (xe_gt_ccs_mode_default(gt))
>> + return xe_sriov_pf_lockdown(xe);
>> +
>> + return 0;
>> +}
>> +
>> +static void gt_finish_ccs_mode_enabling(struct xe_gt *gt)
>> +{
>> + struct xe_device *xe = gt_to_xe(gt);
>> +
>> + if (!IS_SRIOV_PF(xe))
>> + return;
>> +
>> + /* Allow enabling VFs, if CCS-mode changed to default mode */
>> + if (xe_gt_ccs_mode_default(gt))
>> + xe_sriov_pf_end_lockdown(xe);
>> +}
>> +
>> static ssize_t
>> ccs_mode_store(struct device *kdev, struct device_attribute *attr,
>> const char *buff, size_t count)
>> @@ -117,12 +148,6 @@ ccs_mode_store(struct device *kdev, struct device_attribute *attr,
>> u32 num_engines, num_slices;
>> int ret;
>>
>> - if (IS_SRIOV(xe)) {
>> - xe_gt_dbg(gt, "Can't change compute mode when running as %s\n",
>> - xe_sriov_mode_to_string(xe_device_sriov_mode(xe)));
>> - return -EOPNOTSUPP;
>> - }
>> -
>> ret = kstrtou32(buff, 0, &num_engines);
>> if (ret)
>> return ret;
>> @@ -139,21 +164,27 @@ ccs_mode_store(struct device *kdev, struct device_attribute *attr,
>> }
>>
>> /* CCS mode can only be updated when there are no drm clients */
>> - mutex_lock(&xe->drm.filelist_mutex);
>> + guard(mutex)(&xe->drm.filelist_mutex);
>> if (!list_empty(&xe->drm.filelist)) {
>> - mutex_unlock(&xe->drm.filelist_mutex);
>> xe_gt_dbg(gt, "Rejecting compute mode change as there are active drm clients\n");
>> return -EBUSY;
>> }
>>
>> - if (gt->ccs_mode != num_engines) {
>> - xe_gt_info(gt, "Setting compute mode to %d\n", num_engines);
>> - gt->ccs_mode = num_engines;
>> - xe_gt_record_user_engines(gt);
>> - xe_gt_reset(gt);
>> + if (gt->ccs_mode == num_engines)
>> + return count;
>> +
>> + ret = gt_prepare_ccs_mode_enabling(gt);
>> + if (ret) {
>> + xe_gt_dbg(gt, "Rejecting compute mode change as VFs are enabled\n");
>> + return ret;
>> }
>
> The function gt_prepare_ccs_mode_enabling() never returning non-zero
> error if VFs are enabled.
That is not ture, gt_prepapre_ccs_mode_enabling() return
non-zeor(-EBUSY) when vfs enabled.
>
>>
>> - mutex_unlock(&xe->drm.filelist_mutex);
>> + xe_gt_info(gt, "Setting compute mode to %d\n", num_engines);
>> + gt->ccs_mode = num_engines;
>> + xe_gt_record_user_engines(gt);
>> + xe_gt_reset(gt);
>> +
>
> Do we need to apply gt reset after changing CCS mode? If yes, add a
> comment with why do we need it.
>
yes, ensures clean state as we modified new CCS mode for engine setup.
> -Satya.
>
Thanks,
Naresh
>> + gt_finish_ccs_mode_enabling(gt);
>>
>> return count;
>> }
>> diff --git a/drivers/gpu/drm/xe/xe_gt_ccs_mode.h b/drivers/gpu/drm/xe/xe_gt_ccs_mode.h
>> index f8779852cf0d..53a595b0882c 100644
>> --- a/drivers/gpu/drm/xe/xe_gt_ccs_mode.h
>> +++ b/drivers/gpu/drm/xe/xe_gt_ccs_mode.h
>> @@ -20,5 +20,17 @@ static inline bool xe_gt_ccs_mode_enabled(const struct xe_gt *gt)
>> return hweight32(CCS_MASK(gt)) > 1;
>> }
>>
>> +/**
>> + * xe_gt_ccs_mode_default() - Check if CCS mode is default (single CCS mode)
>> + * @gt: GT structure
>> + *
>> + * Return: %true if actual CCS mode is single mode, or
>> + * %false otherwise (CCS in alternate/multi mode)
>> + */
>> +static inline bool xe_gt_ccs_mode_default(struct xe_gt *gt)
>> +{
>> + return gt->ccs_mode == 1;
>> +}
>> +
>> #endif
>>
next prev parent reply other threads:[~2025-12-03 12:43 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-11-28 17:18 [PATCH v1 0/2] drm/xe:Mutual exclusivity between CCS-mode and PF Nareshkumar Gollakoti
2025-11-28 17:18 ` [PATCH v1 1/2] drm/xe: Fix Prevent VFs from exposing the CCS mode sysfs file Nareshkumar Gollakoti
2025-11-28 17:18 ` [PATCH v1 2/2] drm/xe: Mutual exclusivity between CCS-mode and PF Nareshkumar Gollakoti
2025-12-03 8:52 ` K V P, Satyanarayana
2025-12-03 12:13 ` Michal Wajdeczko
2025-12-03 12:43 ` Kumar G, Naresh [this message]
2025-11-28 17:31 ` ✓ CI.KUnit: success for drm/xe:Mutual " Patchwork
2025-11-28 18:46 ` ✓ Xe.CI.BAT: " Patchwork
2025-11-28 19:54 ` ✗ Xe.CI.Full: failure " Patchwork
2025-12-15 5:49 ` Kumar G, Naresh
2026-01-29 17:22 ` [PATCH v2 0/2] " Nareshkumar Gollakoti
2026-01-29 17:22 ` [PATCH v2 1/2] drm/xe: Fix Prevent VFs from exposing the CCS mode sysfs file Nareshkumar Gollakoti
2026-01-29 21:13 ` Michal Wajdeczko
2026-01-30 14:32 ` Kumar G, Naresh
2026-01-30 15:16 ` Rodrigo Vivi
2026-02-06 14:37 ` Kumar G, Naresh
2026-01-29 17:22 ` [PATCH v2 2/2] drm/xe: Mutual exclusivity between CCS-mode and PF Nareshkumar Gollakoti
2026-01-29 21:41 ` Michal Wajdeczko
2026-01-30 14:33 ` Kumar G, Naresh
-- strict thread matches above, loose matches on Subject: below --
2025-11-28 12:38 [V9 PATCH] " Nareshkumar Gollakoti
2025-11-28 17:10 ` [PATCH v1 0/2] " Nareshkumar Gollakoti
2025-11-28 17:10 ` [PATCH v1 2/2] drm/xe: " Nareshkumar Gollakoti
2026-01-15 22:50 ` Michal Wajdeczko
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=b6500ae0-715a-4ee6-9cf8-e5ada3a09dbb@intel.com \
--to=naresh.kumar.g@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=satyanarayana.k.v.p@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