It would be good to know how often #3 happens on a real workload. On February 1, 2014 1:17:10 PM PST, George Spelvin <linux@xxxxxxxxxxx> wrote: >> .. which *does* actually bring up something that might work, and >might >> be a good idea: remove the "restore math state or set TS" from the >> normal kernel paths *entirely*, and move it to the "return to user >> space" phase. > >This definitely seems much more sensible. For processors without >optimized context save/restore, the decision whether to restore eagerly >is best made once the kernel has decided what it's returning *to*. > >From an interrupt handler, there are now three cases: >1) FPU not in use, available for immediate use. >2) FPU in use by user-space; save to <location> before using. >3) FPU in use by kernel code; may not be used (unless we want > to add nested FPU state save hair). -- Sent from my mobile phone. Please pardon brevity and lack of formatting. -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html