[PATCH] kunit: tool: continue past invalid utf-8 output

List overview All Threads
Download

newer

older

[PATCH 0/2] selftest/x86: AMX CPU...

[PATCH] kunit: tool: fix typecheck...

Daniel Latypov

8 Oct 2021 8 Oct '21

9:07 p.m.

kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead: * print out the offending byte as '\x80' * try and continue parsing the output. * as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com --- tools/testing/kunit/kunit.py | 3 ++- tools/testing/kunit/kunit_kernel.py | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/kunit/kunit.py b/tools/testing/kunit/kunit.py index 9c9ed4071e9e..28ae096d4b53 100755 --- a/tools/testing/kunit/kunit.py +++ b/tools/testing/kunit/kunit.py @@ -457,9 +457,10 @@ def main(argv, linux=None): sys.exit(1) elif cli_args.subcommand == 'parse': if cli_args.file == None: + sys.stdin.reconfigure(errors='backslashreplace') kunit_output = sys.stdin else: - with open(cli_args.file, 'r') as f: + with open(cli_args.file, 'r', errors='backslashreplace') as f: kunit_output = f.read().splitlines() request = KunitParseRequest(cli_args.raw_output, None, diff --git a/tools/testing/kunit/kunit_kernel.py b/tools/testing/kunit/kunit_kernel.py index faa6320e900e..f08c6c36a947 100644 --- a/tools/testing/kunit/kunit_kernel.py +++ b/tools/testing/kunit/kunit_kernel.py @@ -135,7 +135,7 @@ class LinuxSourceTreeOperationsQemu(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, - text=True, shell=True) + text=True, shell=True, errors='backslashreplace')

class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): """An abstraction over command line operations performed on a source tree.""" @@ -172,7 +172,7 @@ class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, - text=True) + text=True, errors='backslashreplace')

def get_kconfig_path(build_dir) -> str: return get_file_path(build_dir, KCONFIG_PATH)

base-commit: a032094fc1ed17070df01de4a7883da7bb8d5741

-- 2.33.0.882.g93a45727a2-goog

Show replies by date

Brendan Higgins

8 Oct 8 Oct

9:15 p.m.

On Fri, Oct 8, 2021 at 2:08 PM Daniel Latypov dlatypov@google.com wrote:

...

kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead:

print out the offending byte as '\x80'

try and continue parsing the output.

as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com

Thanks for fixing this!

Reviewed-by: Brendan Higgins brendanhiggins@google.com

Daniel Latypov

11:51 p.m.

On Fri, Oct 8, 2021 at 2:08 PM Daniel Latypov dlatypov@google.com wrote:

...

kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead:

print out the offending byte as '\x80'

try and continue parsing the output.

as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com

tools/testing/kunit/kunit.py | 3 ++- tools/testing/kunit/kunit_kernel.py | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/kunit/kunit.py b/tools/testing/kunit/kunit.py index 9c9ed4071e9e..28ae096d4b53 100755 --- a/tools/testing/kunit/kunit.py +++ b/tools/testing/kunit/kunit.py @@ -457,9 +457,10 @@ def main(argv, linux=None): sys.exit(1) elif cli_args.subcommand == 'parse': if cli_args.file == None:
                  sys.stdin.reconfigure(errors='backslashreplace')

Ugh, pytype doesn't like this even though it's valid. I can squash the error with sys.stdin.reconfigure(errors='backslashreplace') # pytype: disable=attribute-error

I had wanted us to avoid having anything specific to pytype in the code. But mypy (the more common typechecker iirc) hasn't been smart enough to typecheck our code since the QEMU support landed.

If we don't add this directive, both typecheckers will report at least one spurious warning. Should I go ahead and add it, Brendan/David?

...

                    kunit_output = sys.stdin
            else:
                  with open(cli_args.file, 'r') as f:
                  with open(cli_args.file, 'r', errors='backslashreplace') as f:
                          kunit_output = f.read().splitlines()
          request = KunitParseRequest(cli_args.raw_output,
                                      None,
diff --git a/tools/testing/kunit/kunit_kernel.py b/tools/testing/kunit/kunit_kernel.py index faa6320e900e..f08c6c36a947 100644 --- a/tools/testing/kunit/kunit_kernel.py +++ b/tools/testing/kunit/kunit_kernel.py @@ -135,7 +135,7 @@ class LinuxSourceTreeOperationsQemu(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True, shell=True)
                                     text=True, shell=True, errors='backslashreplace')
class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): """An abstraction over command line operations performed on a source tree.""" @@ -172,7 +172,7 @@ class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True)
                                     text=True, errors='backslashreplace')
def get_kconfig_path(build_dir) -> str: return get_file_path(build_dir, KCONFIG_PATH)

base-commit: a032094fc1ed17070df01de4a7883da7bb8d5741

2.33.0.882.g93a45727a2-goog

Daniel Latypov

13 Oct 13 Oct

4:51 p.m.

On Fri, Oct 8, 2021 at 4:51 PM Daniel Latypov dlatypov@google.com wrote:

...

On Fri, Oct 8, 2021 at 2:08 PM Daniel Latypov dlatypov@google.com wrote:

...
kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead:

print out the offending byte as '\x80'

try and continue parsing the output.

as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com

tools/testing/kunit/kunit.py | 3 ++- tools/testing/kunit/kunit_kernel.py | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/kunit/kunit.py b/tools/testing/kunit/kunit.py index 9c9ed4071e9e..28ae096d4b53 100755 --- a/tools/testing/kunit/kunit.py +++ b/tools/testing/kunit/kunit.py @@ -457,9 +457,10 @@ def main(argv, linux=None): sys.exit(1) elif cli_args.subcommand == 'parse': if cli_args.file == None:
                  sys.stdin.reconfigure(errors='backslashreplace')
Ugh, pytype doesn't like this even though it's valid. I can squash the error with sys.stdin.reconfigure(errors='backslashreplace') # pytype: disable=attribute-error

I had wanted us to avoid having anything specific to pytype in the code. But mypy (the more common typechecker iirc) hasn't been smart enough to typecheck our code since the QEMU support landed.

If we don't add this directive, both typecheckers will report at least one spurious warning. Should I go ahead and add it, Brendan/David?

Friendly ping. Should we go ahead and add "# pytype: disable=attribute-error" here?

...

...
                    kunit_output = sys.stdin
            else:
                  with open(cli_args.file, 'r') as f:
                  with open(cli_args.file, 'r', errors='backslashreplace') as f:
                          kunit_output = f.read().splitlines()
          request = KunitParseRequest(cli_args.raw_output,
                                      None,
diff --git a/tools/testing/kunit/kunit_kernel.py b/tools/testing/kunit/kunit_kernel.py index faa6320e900e..f08c6c36a947 100644 --- a/tools/testing/kunit/kunit_kernel.py +++ b/tools/testing/kunit/kunit_kernel.py @@ -135,7 +135,7 @@ class LinuxSourceTreeOperationsQemu(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True, shell=True)
                                     text=True, shell=True, errors='backslashreplace')
class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): """An abstraction over command line operations performed on a source tree.""" @@ -172,7 +172,7 @@ class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True)
                                     text=True, errors='backslashreplace')
def get_kconfig_path(build_dir) -> str: return get_file_path(build_dir, KCONFIG_PATH)

base-commit: a032094fc1ed17070df01de4a7883da7bb8d5741

2.33.0.882.g93a45727a2-goog

Daniel Latypov

20 Oct 20 Oct

11:22 p.m.

On Wed, Oct 13, 2021 at 9:51 AM Daniel Latypov dlatypov@google.com wrote:

...

On Fri, Oct 8, 2021 at 4:51 PM Daniel Latypov dlatypov@google.com wrote:

...
On Fri, Oct 8, 2021 at 2:08 PM Daniel Latypov dlatypov@google.com wrote:

...
kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead:

print out the offending byte as '\x80'

try and continue parsing the output.

as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com

tools/testing/kunit/kunit.py | 3 ++- tools/testing/kunit/kunit_kernel.py | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/kunit/kunit.py b/tools/testing/kunit/kunit.py index 9c9ed4071e9e..28ae096d4b53 100755 --- a/tools/testing/kunit/kunit.py +++ b/tools/testing/kunit/kunit.py @@ -457,9 +457,10 @@ def main(argv, linux=None): sys.exit(1) elif cli_args.subcommand == 'parse': if cli_args.file == None:
                  sys.stdin.reconfigure(errors='backslashreplace')
Ugh, pytype doesn't like this even though it's valid. I can squash the error with sys.stdin.reconfigure(errors='backslashreplace') # pytype: disable=attribute-error

I had wanted us to avoid having anything specific to pytype in the code. But mypy (the more common typechecker iirc) hasn't been smart enough to typecheck our code since the QEMU support landed.

If we don't add this directive, both typecheckers will report at least one spurious warning. Should I go ahead and add it, Brendan/David?
Friendly ping. Should we go ahead and add "# pytype: disable=attribute-error" here?

I've sent out a v2 with this: https://lore.kernel.org/linux-kselftest/20211020232121.1748376-1-dlatypov@go...

...

...
...
                    kunit_output = sys.stdin
            else:
                  with open(cli_args.file, 'r') as f:
                  with open(cli_args.file, 'r', errors='backslashreplace') as f:
                          kunit_output = f.read().splitlines()
          request = KunitParseRequest(cli_args.raw_output,
                                      None,
diff --git a/tools/testing/kunit/kunit_kernel.py b/tools/testing/kunit/kunit_kernel.py index faa6320e900e..f08c6c36a947 100644 --- a/tools/testing/kunit/kunit_kernel.py +++ b/tools/testing/kunit/kunit_kernel.py @@ -135,7 +135,7 @@ class LinuxSourceTreeOperationsQemu(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True, shell=True)
                                     text=True, shell=True, errors='backslashreplace')
class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): """An abstraction over command line operations performed on a source tree.""" @@ -172,7 +172,7 @@ class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True)
                                     text=True, errors='backslashreplace')
def get_kconfig_path(build_dir) -> str: return get_file_path(build_dir, KCONFIG_PATH)

base-commit: a032094fc1ed17070df01de4a7883da7bb8d5741

2.33.0.882.g93a45727a2-goog

Brendan Higgins

25 Oct 25 Oct

9:29 p.m.

On Wed, Oct 13, 2021 at 9:52 AM Daniel Latypov dlatypov@google.com wrote:

...

On Fri, Oct 8, 2021 at 4:51 PM Daniel Latypov dlatypov@google.com wrote:

...
On Fri, Oct 8, 2021 at 2:08 PM Daniel Latypov dlatypov@google.com wrote:

...
kunit.py currently crashes and fails to parse kernel output if it's not fully valid utf-8.

This can come from memory corruption or or just inadvertently printing out binary data as strings.

E.g. adding this line into a kunit test pr_info("\x80") will cause this exception UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 1961: invalid start byte

We can tell Python how to handle errors, see https://docs.python.org/3/library/codecs.html#error-handlers

Unfortunately, it doesn't seem like there's a way to specify this in just one location, so we need to repeat ourselves quite a bit.

Specify `errors='backslashreplace'` so we instead:

print out the offending byte as '\x80'

try and continue parsing the output.

as long as the TAP lines themselves are valid, we're fine.

Signed-off-by: Daniel Latypov dlatypov@google.com

tools/testing/kunit/kunit.py | 3 ++- tools/testing/kunit/kunit_kernel.py | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/kunit/kunit.py b/tools/testing/kunit/kunit.py index 9c9ed4071e9e..28ae096d4b53 100755 --- a/tools/testing/kunit/kunit.py +++ b/tools/testing/kunit/kunit.py @@ -457,9 +457,10 @@ def main(argv, linux=None): sys.exit(1) elif cli_args.subcommand == 'parse': if cli_args.file == None:
                  sys.stdin.reconfigure(errors='backslashreplace')
Ugh, pytype doesn't like this even though it's valid. I can squash the error with sys.stdin.reconfigure(errors='backslashreplace') # pytype: disable=attribute-error

I had wanted us to avoid having anything specific to pytype in the code. But mypy (the more common typechecker iirc) hasn't been smart enough to typecheck our code since the QEMU support landed.

If we don't add this directive, both typecheckers will report at least one spurious warning. Should I go ahead and add it, Brendan/David?
Friendly ping. Should we go ahead and add "# pytype: disable=attribute-error" here?

Sorry, missed this.

Yeah, I am fine with disabling the type checkers if they fail to understand valid code.

...

...
...
                    kunit_output = sys.stdin
            else:
                  with open(cli_args.file, 'r') as f:
                  with open(cli_args.file, 'r', errors='backslashreplace') as f:
                          kunit_output = f.read().splitlines()
          request = KunitParseRequest(cli_args.raw_output,
                                      None,
diff --git a/tools/testing/kunit/kunit_kernel.py b/tools/testing/kunit/kunit_kernel.py index faa6320e900e..f08c6c36a947 100644 --- a/tools/testing/kunit/kunit_kernel.py +++ b/tools/testing/kunit/kunit_kernel.py @@ -135,7 +135,7 @@ class LinuxSourceTreeOperationsQemu(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True, shell=True)
                                     text=True, shell=True, errors='backslashreplace')
class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): """An abstraction over command line operations performed on a source tree.""" @@ -172,7 +172,7 @@ class LinuxSourceTreeOperationsUml(LinuxSourceTreeOperations): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
                                     text=True)
                                     text=True, errors='backslashreplace')
def get_kconfig_path(build_dir) -> str: return get_file_path(build_dir, KCONFIG_PATH)

base-commit: a032094fc1ed17070df01de4a7883da7bb8d5741

2.33.0.882.g93a45727a2-goog

1157

days inactive

1174

days old

linux-kselftest-mirror@lists.linaro.org

5 comments

participants

tags (0)

participants (2)

Brendan Higgins
Daniel Latypov