Re: [PATCH 1/2] drm/xe/pf: Limit fair VF LMEM provisioning

[Michal Wajdeczko <michal.wajdeczko@intel.com>]

On 11.07.2024 22:02, Cavitt, Jonathan wrote:
> -----Original Message-----
> From: Intel-xe <intel-xe-bounces@lists.freedesktop.org> On Behalf Of Michal Wajdeczko
> Sent: Thursday, July 11, 2024 12:23 PM
> To: intel-xe@lists.freedesktop.org
> Cc: Wajdeczko, Michal <Michal.Wajdeczko@intel.com>
> Subject: [PATCH 1/2] drm/xe/pf: Limit fair VF LMEM provisioning
>>
>> Due to the current design of the BO and VRAM manager, any object
>> with XE_BO_FLAG_PINNED flag, which the PF driver uses during VF
>> LMEM provisionining, is created with the TTM_PL_FLAG_CONTIGUOUS
>> flag, which may cause VRAM fragmentation that prevents subsequent
>> allocations of larger objects, like fair VF LMEM provisioning.
>>
>> To avoid such failures, round down fair VF LMEM provisioning size
>> to next power of two size, to compensate what xe_ttm_vram_mgr is
>> doing to achieve contiguous allocations.
>>
>> Fixes: ac6598aed1b3 ("drm/xe/pf: Add support to configure SR-IOV VFs")
>> Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
>> ---
>>  drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c b/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> index db6c213da847..4699b7836001 100644
>> --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> @@ -1543,6 +1543,7 @@ static u64 pf_estimate_fair_lmem(struct xe_gt *gt, unsigned int num_vfs)
>>  	u64 fair;
>>  
>>  	fair = div_u64(available, num_vfs);
>> +	fair = rounddown_pow_of_two(fair);	/* XXX: ttm_vram_mgr & drm_buddy limitation */
>>  	fair = ALIGN_DOWN(fair, alignment);
> 
> If the alignment is also a power of two, then ALIGN_DOWN here probably
> isn't necessary.  But I won't block on it because I'm probably just
> misunderstanding something.

true, it's redundant, but I didn't want to remove it since this new
rounddown_pow_of_two() is (I hope) temporary and I don't want to
accidentally miss mandatory HW requirement

> 
> Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
> -Jonathan Cavitt

thanks!

> 
>>  #ifdef MAX_FAIR_LMEM
>>  	fair = min_t(u64, MAX_FAIR_LMEM, fair);
>> -- 
>> 2.43.0
>>
>>