Hello,
On Wed, 23 Oct 2024 11:41:20 +0200 Benjamin Berg <benja...@sipsolutions.net>
wrote:
> From: Benjamin Berg <benjamin.b...@intel.com>
>
> The PTRACE_GETREGSET API has now existed since Linux 2.6.33. The XSAVE
> CPU feature should also be sufficiently common to be able to rely on it.
>
> With this, define our internal FP state to be the hosts XSAVE data. Add
> discovery for the hosts XSAVE size and place the FP registers at the end
> of task_struct so that we can adjust the size at runtime.
>
> Next we can implement the regset API on top and update the signal
> handling as well as ptrace APIs to use them. Also switch coredump
> creation to use the regset API and finally set HAVE_ARCH_TRACEHOOK.
>
> This considerably improves the signal frames. Previously they might not
> have contained all the registers (i386) and also did not have the
> sizes and magic values set to the correct values to permit userspace to
> decode the frame.
>
> As a side effect, this will permit UML to run on hosts with newer CPU
> extensions (such as AMX) that need even more register state.
I just found kunit starts failing from the mainline commit of this patch on my
qemu-x86 Debian system, as below.
$ git checkout 3f17fed2149192c7d3b76a45a6a87b4ff22cd586
$ ./tools/testing/kunit/kunit.py run --kunitconfig mm/damon/tests/
--build_dir ../kunit.out/
[...]
[22:48:27] Configuring KUnit Kernel ...
Regenerating .config ...
Populating config with:
$ make ARCH=um O=../kunit.out/ olddefconfig
[22:48:30] Building KUnit Kernel ...
Populating config with:
$ make ARCH=um O=../kunit.out/ olddefconfig
Building with:
$ make all compile_commands.json ARCH=um O=../kunit.out/ --jobs=40
[...]
[22:48:46] Starting KUnit Kernel (1/1)...
[22:48:46] ============================================================
Running tests with:
$ ../kunit.out/linux kunit.enable=1 mem=1G console=tty kunit_shutdown=halt
[22:48:46] [ERROR] Test: <missing>: Could not find any KTAP output. Did any
KUnit tests run?
[22:48:46] ============================================================
[22:48:46] Testing complete. Ran 0 tests: errors: 1
[22:48:46] Elapsed time: 19.285s total, 3.805s configuring, 15.475s
building, 0.006s running
>
> Signed-off-by: Benjamin Berg <benjamin.b...@intel.com>
[...]
> -void arch_init_registers(int pid)
> -{
> - struct user_fpxregs_struct fpx_regs;
> - int err;
> -
> - err = ptrace(PTRACE_GETFPXREGS, pid, 0, &fpx_regs);
> - if (!err)
> - return;
> -
> - if (errno != EIO)
> - panic("check_ptrace : PTRACE_GETFPXREGS failed, errno = %d",
> - errno);
> -
> - have_fpx_regs = 0;
> -}
> -#else
> -
> -int get_fp_registers(int pid, unsigned long *regs)
> +int arch_init_registers(int pid)
> {
> - return save_fp_registers(pid, regs);
> + struct iovec iov = {
> + /* Just use plenty of space, it does not cost us anything */
> + .iov_len = 2 * 1024 * 1024,
> + };
> + int ret;
> +
> + iov.iov_base = mmap(NULL, iov.iov_len, PROT_WRITE | PROT_READ,
> + MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
> + if (iov.iov_base == MAP_FAILED)
> + return -ENOMEM;
> +
> + /* GDB has x86_xsave_length, which uses x86_cpuid_count */
> + ret = ptrace(PTRACE_GETREGSET, pid, NT_X86_XSTATE, &iov);
> + if (ret)
> + ret = -errno;
> + munmap(iov.iov_base, 2 * 1024 * 1024);
> +
> + host_fp_size = iov.iov_len;
> +
> + return ret;
> }
And seems it fails from the registers initialization step:
$ ../kunit.out/linux
Core dump limits :
soft - 0
hard - NONE
Checking that ptrace can change system call numbers...OK
Checking syscall emulation for ptrace...OK
Checking environment variables for a tempdir...none found
Checking if /dev/shm is on tmpfs...OK
Checking PROT_EXEC mmap in /dev/shm...OK
Failed to initialize default registers
I'm not familiar with uml code, so reporting this issue first. Any thought,
please?
Thanks,
SJ
[...]
