<zx2c4>
marcan: any plans to switch the kernel to 4k pages?
schuepbs[m] has left #asahi-re [#asahi-re]
<AdryzzOLEDEdition[m]1>
no, the kernel will be kept at 16k so that people start fixing their software. and also, it is faster on 16k
<AdryzzOLEDEdition[m]1>
there should be a 4k patch though
<_jannau_>
zx2c4: no immediate plans. the iommus use a page size of 16k and there is some effort needed to support that with smaller CPU page size. there is wip tree to so some of the work to support that: https://github.com/AsahiLinux/linux/tree/iommu-4k/dev
<zx2c4>
ahh, thanks
<zx2c4>
hmm how do i turn a `add x1, x1, 4095` into a `add x1, x1, 16383` without needing an extra 4 bytes of .text
the_lanetly_052__ has joined #asahi-re
the_lanetly_052 has quit [Ping timeout: 480 seconds]
<_jannau_>
zx2c4: do you have by any chance a register which is guaranteed to have the top 14 bits set? `add x1, x1, xY, lsr #49`
<dougall>
(reminds me of copying decompiled code out of hex-rays and recompiling it with -Oz to make space for binary patches, not that i recommend that)
<sven>
fwiw, i'll pick that 4k iommu back up once robin has finished his iommu api refactoring series. some of the changes become more sane once we already know the iommu instance by the time domain_alloc is called