[PATCH] checkpatch: add warning for avoiding .L prefix symbols in assembly files

[Aditya Srivastava]

Local symbols prefixed with '.L' do not emit symbol table entries, as
they have special meaning for the assembler.

'.L' prefixed symbols can be used within a code region, but should be
avoided for denoting a range of code via 'SYM_*_START/END' annotations.

Add a new check to emit a warning on finding the usage of '.L' symbols
in '.S' files, if it lies within SYM_*_START/END annotation pair.

Suggested-by: Mark Brown <bro...@kernel.org>
Link: https://lore.kernel.org/lkml/2021011221...@sirena.org.uk/
Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
---
scripts/checkpatch.pl | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index 7030c4d6d126..858b5def61e9 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -2501,6 +2501,9 @@ sub process {

my $checklicenseline = 1;

+ # record SYM_*_START/END annotation pair count, for AVOID_L_PREFIX
+ my $sym_start_block = 0;
+
sanitise_line_reset();
my $line;
foreach my $rawline (@rawlines) {
@@ -3590,6 +3593,25 @@ sub process {
}
}

+# check for .L prefix local symbols in .S files
+ if ($realfile =~ /\.S$/) {
+ if ($line =~ /SYM_.*_START/ ||
+ (defined $context_function && $context_function =~ /SYM_.*_START/)) {
+ $sym_start_block++;
+ }
+
+ if ($line=~ /\.L\S+/ && # line contains .L prefixed local symbol
+ $sym_start_block > 0) { # lies between SYM_*_START and SYM_*_END pair
+ WARN("AVOID_L_PREFIX",
+ "Avoid using '.L' prefixed local symbol names for denoting a range of code via 'SYM_*_START/END' annotations; see Documentation/asm-annotations.rst\n" . $herecurr);
+ }
+
+ if ($line =~ /SYM_.*_END/ ||
+ (defined $context_function && $context_function =~ /SYM_.*_END/)) {
+ $sym_start_block--;
+ }
+ }
+
# check we are in a valid source file C or perl if not then ignore this hunk
next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/);

--
2.17.1

[Joe Perches]

On Wed, 2021-01-20 at 12:55 +0530, Aditya Srivastava wrote:
> Local symbols prefixed with '.L' do not emit symbol table entries, as
> they have special meaning for the assembler.
>
> '.L' prefixed symbols can be used within a code region, but should be
> avoided for denoting a range of code via 'SYM_*_START/END' annotations.
>
> Add a new check to emit a warning on finding the usage of '.L' symbols
> in '.S' files, if it lies within SYM_*_START/END annotation pair.

I believe this needs a test for $file as it won't work well on
patches as the SYM_*_START/END may not be in the patch context.

Also, is this supposed to work for local labels like '.L<foo>:'?
I don't think a warning should be generated for those.

[Aditya]

On 20/1/21 2:51 pm, Joe Perches wrote:
> On Wed, 2021-01-20 at 12:55 +0530, Aditya Srivastava wrote:
>> Local symbols prefixed with '.L' do not emit symbol table entries, as
>> they have special meaning for the assembler.
>>
>> '.L' prefixed symbols can be used within a code region, but should be
>> avoided for denoting a range of code via 'SYM_*_START/END' annotations.
>>
>> Add a new check to emit a warning on finding the usage of '.L' symbols
>> in '.S' files, if it lies within SYM_*_START/END annotation pair.
>
> I believe this needs a test for $file as it won't work well on
> patches as the SYM_*_START/END may not be in the patch context.
>

Okay.

> Also, is this supposed to work for local labels like '.L<foo>:'?
> I don't think a warning should be generated for those.
>

Yes, currently it will generate warning for all symbols which start
with .L and have non- white character symbol following it, if it is
lying within SYM_*_START/END annotation pair.

Should I reduce the check to \.L_\S+ instead? (please note "_"
following ".L")
Pardon me, I'm not good with assembly :/

Thanks
Aditya

[Joe Perches]

Use grep first. That would still match several existing labels.

> Pardon me, I'm not good with assembly :/

Spending time reading docs can help with that.

Mark? Can you please comment about the below?

I believe the test should be:

if ($realfile =~ /\.S$/ &&
$line =~ /^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {
WARN(...);
}

so that only this code currently matches:

$ git grep -P '^\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L' -- '*.S'
arch/x86/boot/compressed/head_32.S:SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated)
arch/x86/boot/compressed/head_32.S:SYM_FUNC_END(.Lrelocated)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_END(.Lrelocated)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_START_LOCAL_NOALIGN(.Lpaging_enabled)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_END(.Lpaging_enabled)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_START_LOCAL_NOALIGN(.Lno_longmode)
arch/x86/boot/compressed/head_64.S:SYM_FUNC_END(.Lno_longmode)
arch/x86/boot/pmjump.S:SYM_FUNC_START_LOCAL_NOALIGN(.Lin_pm32)
arch/x86/boot/pmjump.S:SYM_FUNC_END(.Lin_pm32)
arch/x86/entry/entry_64.S:SYM_CODE_START_LOCAL_NOALIGN(.Lbad_gs)
arch/x86/entry/entry_64.S:SYM_CODE_END(.Lbad_gs)
arch/x86/lib/copy_user_64.S:SYM_CODE_START_LOCAL(.Lcopy_user_handle_tail)
arch/x86/lib/copy_user_64.S:SYM_CODE_END(.Lcopy_user_handle_tail)
arch/x86/lib/getuser.S:SYM_CODE_START_LOCAL(.Lbad_get_user_clac)
arch/x86/lib/getuser.S:SYM_CODE_END(.Lbad_get_user_clac)
arch/x86/lib/getuser.S:SYM_CODE_START_LOCAL(.Lbad_get_user_8_clac)
arch/x86/lib/getuser.S:SYM_CODE_END(.Lbad_get_user_8_clac)
arch/x86/lib/putuser.S:SYM_CODE_START_LOCAL(.Lbad_put_user_clac)
arch/x86/lib/putuser.S:SYM_CODE_END(.Lbad_put_user_clac)
arch/x86/realmode/rm/wakeup_asm.S:SYM_DATA_START_LOCAL(.Lwakeup_idt)
arch/x86/realmode/rm/wakeup_asm.S:SYM_DATA_END(.Lwakeup_idt)

[Nick Desaulniers]

Josh, I assume objtool does not annotate code under:
arch/x86/boot/
arch/x86/entry/
arch/x86/realmode/
?

The rest, ie
arch/x86/lib/
seem potentially relevant?
--
Thanks,
~Nick Desaulniers

[Josh Poimboeuf]

Both arch/x86/entry/* and arch/x86/lib/* are read by objtool and should
probably be fixed up.

--
Josh

[Aditya]

Joe, with this regex, we won't be able to match
"jmp .L_restore
SYM_FUNC_END(\name)"

which was replaced in this patch in the discussion:
https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/phKe-Tb6CgAJ

Here, "jmp .L_restore" was also replaced to fix the error.

However, if this
regex(/^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) is
correct, maybe we don't need to check for $file as we are now checking
for just one line.

What do you think?

Thanks
Aditya

[Joe Perches]

On Fri, 2021-01-22 at 18:48 +0530, Aditya wrote:
> On 21/1/21 12:13 am, Joe Perches wrote:
> > I believe the test should be:
> >
> > if ($realfile =~ /\.S$/ &&
> > $line =~ /^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {
>
> Joe, with this regex, we won't be able to match
> "jmp .L_restore
> SYM_FUNC_END(\name)"

I think that's not an issue.

> which was replaced in this patch in the discussion:
> https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/phKe-Tb6CgAJ
>
> Here, "jmp .L_restore" was also replaced to fix the error.

Notice that this line was also replaced in the same patch:

#ifdef CONFIG_PREEMPTION
-SYM_CODE_START_LOCAL_NOALIGN(.L_restore)
+SYM_CODE_START_LOCAL_NOALIGN(__thunk_restore)

> However, if this
> regex(/^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) is
> correct, maybe we don't need to check for $file as we are now checking
> for just one line.
>
> What do you think?

The test I wrote was complete, did not use $file and emits a
warning on the use of CODE_SYM_START_LOCAL_NOALIGN(.L_restore)

I think it's sufficient.

[Aditya]

Okay, Thanks.. I will send the modified patch :)

Thanks
Aditya

[Fāng-ruì Sòng]

I am slightly concerned with the direction here. jmp .Lsym is fine
and is probably preferable in cases where you absolutely want to emit
a local symbol.
(there is a related -z unique-symbol GNU ld+FGKASLR thing I shall
analyze in my spare time)

If the problem is just that STT_SECTION symbols are missing, can
objtool do something to conceptually synthesize them?

[Aditya Srivastava]

objtool requires that all code must be contained in an ELF symbol.
Symbol names that have a '.L' prefix do not emit symbol table entries, as

they have special meaning for the assembler.

'.L' prefixed symbols can be used within a code region, but should be
avoided for denoting a range of code via 'SYM_*_START/END' annotations.

Add a new check to emit a warning on finding the usage of '.L' symbols

for '.S' files in arch/x86/entry/* and arch/x86/lib/*, if it denotes
range of code via SYM_*_START/END annotation pair.

Suggested-by: Mark Brown <bro...@kernel.org>
Link: https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/4staOlf-CgAJ

Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
---

* Applies perfectly on next-20210122

Changes in v2:
- Reduce the check to only SYM_*_START/END lines
- Reduce the check for only .S files in arch/x86/entry/* and arch/x86/lib/* as suggested by Josh and Nick
- Modify commit message

scripts/checkpatch.pl | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index 7030c4d6d126..e36cdf96dfe3 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -3590,6 +3590,13 @@ sub process {

}
}

+# check for .L prefix local symbols in .S files

+ if ($realfile =~ m@^arch/x86/(?:entry|lib)/.*\.S$@ &&
+ $line =~ /^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {

+ WARN("AVOID_L_PREFIX",
+ "Avoid using '.L' prefixed local symbol names for denoting a range of code via 'SYM_*_START/END' annotations; see Documentation/asm-annotations.rst\n" . $herecurr);
+ }
+

[Joe Perches]

On Sat, 2021-01-23 at 20:44 +0530, Aditya Srivastava wrote:
> objtool requires that all code must be contained in an ELF symbol.
> Symbol names that have a '.L' prefix do not emit symbol table entries, as
> they have special meaning for the assembler.
>
> '.L' prefixed symbols can be used within a code region, but should be
> avoided for denoting a range of code via 'SYM_*_START/END' annotations.
>
> Add a new check to emit a warning on finding the usage of '.L' symbols
> for '.S' files in arch/x86/entry/* and arch/x86/lib/*, if it denotes
> range of code via SYM_*_START/END annotation pair.
>
> Suggested-by: Mark Brown <bro...@kernel.org>
> Link: https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/4staOlf-CgAJ

Please do not use groups.google.com links, or if you must, please
use links that are readable.

For instance, this is a better link as it shows the context without
struggling with the poor interface:

https://groups.google.com/g/clang-built-linux/c/E-naBMt_1SM

> Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
> ---
> * Applies perfectly on next-20210122
>
> Changes in v2:
> - Reduce the check to only SYM_*_START/END lines
> - Reduce the check for only .S files in arch/x86/entry/* and arch/x86/lib/* as suggested by Josh and Nick

I think that's unnecessary.

> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl

[]
> @@ -3590,6 +3590,13 @@ sub process {
>   }
>   }
>  
>
> +# check for .L prefix local symbols in .S files
> + if ($realfile =~ m@^arch/x86/(?:entry|lib)/.*\.S$@ &&

Using /.S$/ should be enough

> + $line =~ /^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {

This might need to be

+ $line =~ /^\+\s*(?:[A-Z]+_)?SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {

[Aditya]

On 23/1/21 10:51 pm, Joe Perches wrote:
> On Sat, 2021-01-23 at 20:44 +0530, Aditya Srivastava wrote:
>> objtool requires that all code must be contained in an ELF symbol.
>> Symbol names that have a '.L' prefix do not emit symbol table entries, as
>> they have special meaning for the assembler.
>>
>> '.L' prefixed symbols can be used within a code region, but should be
>> avoided for denoting a range of code via 'SYM_*_START/END' annotations.
>>
>> Add a new check to emit a warning on finding the usage of '.L' symbols
>> for '.S' files in arch/x86/entry/* and arch/x86/lib/*, if it denotes
>> range of code via SYM_*_START/END annotation pair.
>>
>> Suggested-by: Mark Brown <bro...@kernel.org>
>> Link: https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/4staOlf-CgAJ
>
> Please do not use groups.google.com links, or if you must, please
> use links that are readable.
>
> For instance, this is a better link as it shows the context without
> struggling with the poor interface:
>
> https://groups.google.com/g/clang-built-linux/c/E-naBMt_1SM
>

Okay, Got it.. I'll use lkml link for the best.

>> Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
>> ---
>> * Applies perfectly on next-20210122
>>
>> Changes in v2:
>> - Reduce the check to only SYM_*_START/END lines
>> - Reduce the check for only .S files in arch/x86/entry/* and arch/x86/lib/* as suggested by Josh and Nick
>
> I think that's unnecessary.
>
>> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> []
>> @@ -3590,6 +3590,13 @@ sub process {
>>   }
>>   }
>>  
>>
>> +# check for .L prefix local symbols in .S files
>> + if ($realfile =~ m@^arch/x86/(?:entry|lib)/.*\.S$@ &&
>
> Using /.S$/ should be enough
>
>> + $line =~ /^\+\s*SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {
>
> This might need to be
>
> + $line =~ /^\+\s*(?:[A-Z]+_)?SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {
>

Okay.. Thanks. I'll modify the patch accordingly.

Thanks
Aditya

[Aditya Srivastava]

objtool requires that all code must be contained in an ELF symbol.
Symbol names that have a '.L' prefix do not emit symbol table entries, as
they have special meaning for the assembler.

'.L' prefixed symbols can be used within a code region, but should be
avoided for denoting a range of code via 'SYM_*_START/END' annotations.

Add a new check to emit a warning on finding the usage of '.L' symbols

for '.S' files, if it denotes range of code via SYM_*_START/END

annotation pair.

Suggested-by: Mark Brown <bro...@kernel.org>

Link: https://lore.kernel.org/lkml/2021011221...@sirena.org.uk

Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
---
* Applies perfectly on next-20210122

Changes in v3:
- Modify regex for SYM_*_START/END pair
- remove check for arch/x86/entry/* and arch/x86/lib/*
- change 'Link:' in commit message to lkml
- Modify commit description accordingly

Changes in v2:
- Reduce the check to only SYM_*_START/END lines
- Reduce the check for only .S files in arch/x86/entry/* and arch/x86/lib/* as suggested by Josh and Nick

- Modify commit message

scripts/checkpatch.pl | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl

index 7030c4d6d126..4a03326c87b6 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl

@@ -3590,6 +3590,13 @@ sub process {
}
}

+# check for .L prefix local symbols in .S files

+ if ($realfile =~ /\.S$/ &&

+ $line =~ /^\+\s*(?:[A-Z]+_)?SYM_[A-Z]+_(?:START|END)(?:_[A-Z_]+)?\s*\(\s*\.L/) {

+ WARN("AVOID_L_PREFIX",
+ "Avoid using '.L' prefixed local symbol names for denoting a range of code via 'SYM_*_START/END' annotations; see Documentation/asm-annotations.rst\n" . $herecurr);
+ }
+

[Joe Perches]

On Sun, 2021-01-24 at 00:34 +0530, Aditya Srivastava wrote:
> objtool requires that all code must be contained in an ELF symbol.
> Symbol names that have a '.L' prefix do not emit symbol table entries, as
> they have special meaning for the assembler.
>
> '.L' prefixed symbols can be used within a code region, but should be
> avoided for denoting a range of code via 'SYM_*_START/END' annotations.
>
> Add a new check to emit a warning on finding the usage of '.L' symbols
> for '.S' files, if it denotes range of code via SYM_*_START/END
> annotation pair.
>
> Suggested-by: Mark Brown <bro...@kernel.org>
> Link: https://lore.kernel.org/lkml/2021011221...@sirena.org.uk
> Signed-off-by: Aditya Srivastava <yashs...@gmail.com>

Acked-by: Joe Perches <j...@perches.com>

[Nick Desaulniers]

On Sat, Jan 23, 2021 at 1:01 PM Joe Perches <j...@perches.com> wrote:
>
> On Sun, 2021-01-24 at 00:34 +0530, Aditya Srivastava wrote:
> > objtool requires that all code must be contained in an ELF symbol.
> > Symbol names that have a '.L' prefix do not emit symbol table entries, as
> > they have special meaning for the assembler.
> >
> > '.L' prefixed symbols can be used within a code region, but should be
> > avoided for denoting a range of code via 'SYM_*_START/END' annotations.
> >
> > Add a new check to emit a warning on finding the usage of '.L' symbols
> > for '.S' files, if it denotes range of code via SYM_*_START/END
> > annotation pair.
> >
> > Suggested-by: Mark Brown <bro...@kernel.org>
> > Link: https://lore.kernel.org/lkml/2021011221...@sirena.org.uk
> > Signed-off-by: Aditya Srivastava <yashs...@gmail.com>
>
> Acked-by: Joe Perches <j...@perches.com>

Acked-by: Nick Desaulniers <ndesau...@google.com>

Thanks for the patch Aditya!

--
Thanks,
~Nick Desaulniers