ChanServ changed the topic of #armlinux to: ARM kernel talk [Upstream kernel, find your vendor forums for questions about their kernels] | https://libera.irclog.whitequark.org/armlinux
mraynal has quit [Remote host closed the connection]
mraynal has joined #armlinux
gshan has quit [Ping timeout: 276 seconds]
lain6141 has quit [Read error: Connection reset by peer]
headless has quit [Read error: Connection reset by peer]
mripard has joined #armlinux
dmart has joined #armlinux
<arnd>
maz, vireshk: I think found what's going on with tlbi, see CONFIG_ARM64_TLB_RANGE: "depends on AS_HAS_ARMV8_4"
<vireshk>
arnd, "support when binutils >= 2.30" ?
<arnd>
vireshk: exactly. I using binutils-2.30 works fine with the new code, older versions need the same compile-time check for CONFIG_ARM64_TLB_RANGE with a fallback
<vireshk>
makes sense, thanks
<arnd>
the easiest way is probably to just make KVM itself depend on CONFIG_ARM64_TLB_RANGE and force users to upgrade to a sensible toolchain version if they care. maz may have an idea for making that option a little more fine-grained, but 2.30 is already really old
<arnd>
I see the oldest compiler we support in theory is gcc-5.1 from early 2015, which roughly corresponds to binutils-2.25
<maz>
arnd: yup, came to the same conclusion. I'll add a dependency on AS_HAS_ARMV8_4, and that will quickly weed out the problems.
<maz>
hopefully one day, we'll bump up the binutils (and compiler) requirements to something from this decade.
<arnd>
we were really close to changing it to gcc-8 recently, but torvalds in the end decided to stay with gcc-5
<arnd>
I think we are overdue for raising the minimum toolchain. There is only one more merge window before the next LTS kernel, so we may as well wait until that is done and raise the minimum for 6.13
<arnd>
rhel8 and its clones ship with gcc-8 and binutils-2.30, so that would be a likely candidate. Everything else that matters seems to have newer versions than this, e.g. Debian 10 is already EOL and has binutils-2.31