Hi, On Fri, Feb 2, 2024 at 11:43 AM Mark Brown <broonie@xxxxxxxxxx> wrote: > > On Fri, Feb 02, 2024 at 06:07:54PM +0000, Dave Martin wrote: > > > So, if the only reason for trying to migrate to vmalloc() is to cope > > with an insanely sized regset on arm64, I think somehow or other we can > > avoid that. > > With SME we do routinely see the full glory of the 64K regset for ZA in > emulated systems so I think we have to treat it as an issue. Ah, got it. 64K is much less likely to be as big of a problem (only an order 4 allocation), but you're right that it's still a size where kvmalloc() would be an improvement. With that in mind I'll plan to send out a v2 of my patch where I use kvmalloc() instead of vmalloc() and update the commit description a bit, including a link to this thread. Then I will assume that others on this thread will move forward with actually making the allocations smaller. Please yell if the above sounds wrong. :-) -Doug
