A few improvements for riscv

[AlexGhiti]

This reenables KASAN_INLINE and sets 4-level page table by default: note that we can use a kernel command line parameter instead if you prefer ("no5lvl").

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[a-nogikh]

Thanks!

For the dashboard/config/linux/bits: reenable KASAN_INLINE for riscv64 commit, you also need to regenerate the configs:

syzkaller/dashboard/config/linux/README.md

Lines 112 to 119 in 7cc1700

	To update kernel configs:
	- change config fragments as necessary (e.g. add additional configs to [subsystems.yml](bits/subsystems.yml) along
	with minimal kernel version)
	- run `syz-env make configs SOURCEDIR=/path/to/existing/linux/checkout`
	(note: it will be mrproper-ed and a number of remotes will be added)
	(see [this](/docs/contributing.md#using-syz-env) on how to setup/use `syz-env`)
	- check in config fragments and changed kernel configs and send a PR
	- changes will be deployed to `syzbot` within a day after merging

I.e. run ./tools/syz-env make configs SOURCEDIR=/path/to/linux/checkout and commit the changes.

And regarding vm/qemu: Run riscv64 kernel using 4-level page table: CI checks complain about the capitalization of Run, it should have been titled vm/qemu: run riscv64 kernel using 4-level page table.

[AlexGhiti]

Arf, sorry about the "Run"...

Since I'm here, I noticed riscv is excluded from a few configs in files like dashboard/config/linux/bits/subsystems.yml although we implement them and they appear in the final generated config. Is there something to do here? Not sure I understood the goal of those.

Thanks for your quick answer!

[a-nogikh]

I noticed riscv is excluded from a few configs in files like dashboard/config/linux/bits/subsystems.yml although we implement them and they appear in the final generated config. Is there something to do here? Not sure I understood the goal of those.

Overall, we try to reduce the number of enabled configs for the platforms that we run in emulation (like riscv) - the execution is very slow and we'd rather focus it on the parts more relevant to the specific platform rather than on the generic code that's already well tested elsewhere. E.g. see the reduced bit usage here:

https://github.com/google/syzkaller/blob/7cc17001e974d9ff7d7ce042c4e2ad2e8ec0e55e/dashboard/config/linux/main.yml#L24C88-L24C95

Specifically in dashboard/config/linux/bits/subsystems.yml, I guess that the reasons were one of:

• The intent similar to what I described above (we just did not use the -reduced condition instead, but should have done so).
• We cannot fuzz the feature yet (no descriptions/not supported by the emulator).
• The kernel was failing to build/boot with these enabled (though for these we usually leave an explicit comment).

If you believe that we'd better still fuzz some of them on the riscv syzbot instance, feel free to send a PR to switch them on or just let us know :)

[AlexGhiti]

@a-nogikh Sorry to ping you, I'm not sure you were notified by my force-push. Or maybe you ignored it because of the ci/build failure? In that case, I would definitely need help to understand it :) Thanks

[a-nogikh]

Hi @AlexGhiti !
Feel free to write PTAL or explicitly re-request review in such cases :) Sometimes it's hard to distinguish the force pushes that we all while working on the PR and when the PR actually needs to be reviewed again.

Regarding the CI failures -- could you please rebase your changes on top of the latest master? We recently did fix some syzkaller bugs that could have led to the failures like the ones you got.

[a-nogikh]

Filed #5775.
Let's try again.

[AlexGhiti]

Thanks @a-nogikh!

[a-nogikh]

Thanks for preparing the PR :)