commit 7eb8c8fb29fd8b6620c66f502bbed182f1b6c5c4
author Marco Vanotti <mvanotti@google.com> Tue Apr 28 02:39:21 2020 +0000
committer CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Tue Apr 28 02:39:21 2020 +0000
tree 75aac34e9842ead31d603a135a803750c8d8f267
parent 102bedf5e86861d3fd9009783193f90562be6eca

[zircon][debugger] Don't write reserved part of mxcsr register.

This CL fixes a kernel panic in zircon, in which the
`zx_thread_write_state` syscall could overwrite the entire `mxcsr`,
including the reserved bits.

According to section "10.2.3 MXCSR Control and Status Register" in
Intel's Software Developer Manual Volume 1: Basic Architecture,
writing a non-zero value to the reserved bits causes a general
protection fault.

The fix reads the mxcsr_mask from the `x86_xsave_legacy_area` and checks
whether the function is trying to write into a non-set bit. Note that
it's possible for the mxcsr_mask to be zero, and in that case, the mask
is hard-coded to the value 0x0000ffbf, as specified by the intel and amd
manuals[0][1].

If the mxcsr has invalid bits, the function returns `ZX_ERR_INVALID_ARGS`,
not copying anything to the fxsave area.

I added a unit test to test this behavior: writing all ones to the
register would trigger a general protection fault if this fix is not included.

To run the test:

```
$ fx set bringup.x64 \
    --with-base //garnet/packages/tests:zircon \
    --with-base //bundles/bringup:tests
```

And run it with `runtests -t threads-test`

BUG=50632
This vulnerability was reported by: Quarkslab

[0]: Intel 64 and IA-32 Architectures Software Developer’s Manual
       Volume 1: Basic Architecture
       Section: 11.6.6 Guidelines for Writing to the MXCSR Register
[1]: AMD64 Architecture Programmer’s Manual
       Volume 2: System Programming
       Section: 11.5.9  MXCSR State Management

Change-Id: If1794258a086e20beb24591e63d8da87b74b828a
Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/383146
Commit-Queue: Marco Vanotti <mvanotti@google.com>
Reviewed-by: Gianfranco Valentino <gevalentino@google.com>
Reviewed-by: Venkatesh Srinivas <venkateshs@google.com>
Reviewed-by: Brett Wilson <brettw@google.com>
Testability-Review: Gianfranco Valentino <gevalentino@google.com>
Testability-Review: Brett Wilson <brettw@google.com>