Re: [PATCH] drm/i915: replace in_atomic() with manually set flag

2025-02-03 Thread Maciej Patelczyk
On 28.01.2025 10:29, Krzysztof Karas wrote: Hi Maciej, The locked==true looks OK. thanks for review! However, I fear that there is some corner case with locked==false. 1 or 2 calls back in chain looks good. CI failures needs to be analyzed. Yup, I already did that recently. I thought th

Re: [PATCH] drm/i915: replace in_atomic() with manually set flag

2025-01-28 Thread Krzysztof Karas
Hi Maciej, > The locked==true looks OK. > thanks for review! > However, I fear that there is some corner case with locked==false. 1 or 2 > calls back in chain looks good. > > CI failures needs to be analyzed. > Yup, I already did that recently. I thought those were suspicious, but I could not r

Re: [PATCH] drm/i915: replace in_atomic() with manually set flag

2025-01-23 Thread Maciej Patelczyk
On 10.01.2025 16:46, Krzysztof Karas wrote: GuC code uses in_atomic() function to determine if the current context is atomic. As noted by the function's description it should not be used in driver code, as it is not guaranteed to determine atomicity correctly in every instance. This is also poin

[PATCH] drm/i915: replace in_atomic() with manually set flag

2025-01-10 Thread Krzysztof Karas
GuC code uses in_atomic() function to determine if the current context is atomic. As noted by the function's description it should not be used in driver code, as it is not guaranteed to determine atomicity correctly in every instance. This is also pointed out by the FIXME note suggesting that the c