checkpatch: add warning for avoiding .L prefix symbols in assembly files
PATCH 1/1 checkpatch: add warning for avoiding .L prefix symbols in assembly files
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 warning on finding the usage of '.L' symbols in '.S' files.
--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)
scripts/checkpatch.pl +6 -0
@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/);
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 35 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
'.L' prefixed symbols can be used within a code region, but should be avoided for denoting a range of code via 'SYM_*_START/END' annotations.
Add a new check to emit warning on finding the usage of '.L' symbols in '.S' files.
Suggested-by Mark Brown <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)scripts/checkpatch.pl +6
-1@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/); --
Quoted 1 line
2.17.1
From an earlier conversation:
Quoted 3 lines
So basically, you can use an .L symbol *inside* a function or a code segment, you just can't use the .L symbol to contain the code using a SYM_*_START/END annotation pair.
So this check warns on all uses of the .L prefix
I think that might be incorrect.
Thanks,
Dwaipayan.
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 39 lines · Dwaipayan Ray
Quoted 35 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
'.L' prefixed symbols can be used within a code region, but should be avoided for denoting a range of code via 'SYM_*_START/END' annotations.
Add a new check to emit warning on finding the usage of '.L' symbols in '.S' files.
Suggested-by Mark Brown <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)scripts/checkpatch.pl +6
-1@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/); --Quoted 1 line
2.17.1
From an earlier conversation:
Quoted 7 lines
Quoted 3 lines
So basically, you can use an .L symbol *inside* a function or a code segment, you just can't use the .L symbol to contain the code using a SYM_*_START/END annotation pair.
So this check warns on all uses of the .L prefix
I think that might be incorrect.
Hey Dwaipayan,
I think you missed this:
Quoted 5 lines
- If the line contains ".L" prefixed symbol, give user a warning/check, so that they can ensure that the line is not inside START/END block. (As we may not be able to make sure about the same, if the START/END line is not in the patch; otherwise we could run a while loop)
At best, I think, we could use $context_function, which should work for patches, but again, this will not work for files.
Do you have any suggestions?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 45 lines · Aditya
Quoted 43 lines · Dwaipayan Ray
Quoted 35 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
'.L' prefixed symbols can be used within a code region, but should be avoided for denoting a range of code via 'SYM_*_START/END' annotations.
Add a new check to emit warning on finding the usage of '.L' symbols in '.S' files.
Suggested-by Mark Brown <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)scripts/checkpatch.pl +6
-1@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/); --Quoted 1 line
2.17.1
From an earlier conversation:
Quoted 3 lines
So basically, you can use an .L symbol *inside* a function or a code segment, you just can't use the .L symbol to contain the code using a SYM_*_START/END annotation pair.
Quoted 13 lines
Quoted 3 lines
So this check warns on all uses of the .L prefix
I think that might be incorrect.Hey Dwaipayan,
I think you missed this:Quoted 5 lines
- If the line contains ".L" prefixed symbol, give user a warning/check, so that they can ensure that the line is not inside START/END block. (As we may not be able to make sure about the same, if the START/END line is not in the patch; otherwise we could run a while loop)
Then again what percentage of users actually intended to use it in a local scope? That is actually a correct usage and checkpatch will keep complaining - there will be no way to remove the warning.
Quoted 5 lines
At best, I think, we could use $context_function, which should work for patches, but again, this will not work for files.
Do you have any suggestions?
Does $context_function work for .S files? Am not sure. But again that will not help much.
The main concern is not to allow .L prefix within a SYM_*_START...SYM_*_END annotation pair.
That means something like this will be disallowed:
SYM_FUNC_START(memset)
.L ... etc
SYM_FUNC_END(memset)
One possible way can be to extract the context after getting a SYM_*_START. Following which the lines could be checked for a .L preifx.
But complexity wise I dont know if it would be worth it.
Thanks & Regards,
Dwaipayan.
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 72 lines · Dwaipayan Ray
Quoted 45 lines · Aditya
Quoted 43 lines · Dwaipayan Ray
Quoted 35 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
'.L' prefixed symbols can be used within a code region, but should be avoided for denoting a range of code via 'SYM_*_START/END' annotations.
Add a new check to emit warning on finding the usage of '.L' symbols in '.S' files.
Suggested-by Mark Brown <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)scripts/checkpatch.pl +6
-1@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/); --Quoted 1 line
2.17.1
From an earlier conversation:
Quoted 3 lines
So basically, you can use an .L symbol *inside* a function or a code segment, you just can't use the .L symbol to contain the code using a SYM_*_START/END annotation pair.
Quoted 13 lines
Quoted 3 lines
So this check warns on all uses of the .L prefix
I think that might be incorrect.Hey Dwaipayan,
I think you missed this:Quoted 5 lines
- If the line contains ".L" prefixed symbol, give user a warning/check, so that they can ensure that the line is not inside START/END block. (As we may not be able to make sure about the same, if the START/END line is not in the patch; otherwise we could run a while loop)
Then again what percentage of users actually intended to use it in a local scope? That is actually a correct usage and checkpatch will keep complaining - there will be no way to remove the warning.
Quoted 5 lines
At best, I think, we could use $context_function, which should work for patches, but again, this will not work for files.
Do you have any suggestions?
Does $context_function work for .S files? Am not sure.> But again that will not help much.
Quoted 9 lines
The main concern is not to allow .L prefix within a SYM_*_START...SYM_*_END annotation pair.
That means something like this will be disallowed:
SYM_FUNC_START(memset).L ... etc
SYM_FUNC_END(memset)
Yes, $context_function works inside .S files.
So, I think if it is inside one of the SYM_*_START contexts, it should work for us.
Quoted 5 lines
One possible way can be to extract the context after getting a SYM_*_START. Following which the lines could be checked for a .L preifx.
Quoted 1 line
But complexity wise I dont know if it would be worth it.
I think, we could do this for files.
But I am not sure about the complexity part.
Lukas, what do you think?
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 88 lines · Aditya
Quoted 72 lines · Dwaipayan Ray
Quoted 45 lines · Aditya
Quoted 43 lines · Dwaipayan Ray
Quoted 35 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
'.L' prefixed symbols can be used within a code region, but should be avoided for denoting a range of code via 'SYM_*_START/END' annotations.
Add a new check to emit warning on finding the usage of '.L' symbols in '.S' files.
Suggested-by Mark Brown <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)scripts/checkpatch.pl +6
-1@@ -3590,6 +3590,12 @@ sub process { } } +# check for .L prefix local symbols in .S files + if ($realfile =~ /\.S$/ && $line =~ /\.L\S+/) { + 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" . $herecurr); + } + # check we are in a valid source file C or perl if not then ignore this hunk next if ($realfile !~ /\.(h|c|pl|dtsi|dts)$/); --Quoted 1 line
2.17.1
From an earlier conversation:
Quoted 3 lines
So basically, you can use an .L symbol *inside* a function or a code segment, you just can't use the .L symbol to contain the code using a SYM_*_START/END annotation pair.
Quoted 13 lines
Quoted 3 lines
So this check warns on all uses of the .L prefix
I think that might be incorrect.Hey Dwaipayan,
I think you missed this:Quoted 5 lines
- If the line contains ".L" prefixed symbol, give user a warning/check, so that they can ensure that the line is not inside START/END block. (As we may not be able to make sure about the same, if the START/END line is not in the patch; otherwise we could run a while loop)
Then again what percentage of users actually intended to use it in a local scope? That is actually a correct usage and checkpatch will keep complaining - there will be no way to remove the warning.
Quoted 5 lines
At best, I think, we could use $context_function, which should work for patches, but again, this will not work for files.
Do you have any suggestions?
Does $context_function work for .S files? Am not sure.> But again that will not help much.
Quoted 9 lines
The main concern is not to allow .L prefix within a SYM_*_START...SYM_*_END annotation pair.
That means something like this will be disallowed:
SYM_FUNC_START(memset).L ... etc
SYM_FUNC_END(memset)Yes, $context_function works inside .S files.
So, I think if it is inside one of the SYM_*_START contexts, it should work for us.
Quoted 8 lines
Quoted 5 lines
One possible way can be to extract the context after getting a SYM_*_START. Following which the lines could be checked for a .L preifx.
Quoted 1 line
But complexity wise I dont know if it would be worth it.
I think, we could do this for files.
Just run it on the current kernel source and find out how many cases would generally match.
If it is just a few hundred occurrences, use a simple rule, not a complex one that hardly works on patches.
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH 1/1 checkpatch: add warning for avoiding .L prefix symbols in assembly files
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 warning on finding the usage of '.L' symbols in '.S' files, if it lies within SYM_*_START/END annotation pair.
--- scripts/checkpatch.pl | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
scripts/checkpatch.pl +22 -0
@@ -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)$/);
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 8 lines · 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 warning on finding the usage of '.L' symbols
to emit a warning
Quoted 3 lines
in '.S' files, if it lies within SYM_*_START/END annotation pair.
Suggested-by Mark Brown <[email protected]>
Hmm, that is not an email?
Did you test running it on the patch that introduced the bug initially which then led to the overall discussion I pointed out to you?
Lukas
Quoted 47 lines
Signed-off-by Aditya Srivastava <[email protected]>--- scripts/checkpatch.pl | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)scripts/checkpatch.pl +22
-1@@ -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)$/); --
Quoted 2 lines
2.17.1
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 5 lines · Lukas Bulwahn
Quoted 4 lines · Aditya Srivastava
Local symbols prefixed with '.L' do not emit symbol table entries, as they have special meaning for the assembler.
Quoted 7 lines
Quoted 4 lines
'.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 warning on finding the usage of '.L' symbols
to emit a warning
Okay
Quoted 6 lines
Quoted 3 lines
in '.S' files, if it lies within SYM_*_START/END annotation pair.
Suggested-by Mark Brown <[email protected]>Hmm, that is not an email?
Oops, I'll fix it.
Quoted 3 lines
Did you test running it on the patch that introduced the bug initially which then led to the overall discussion I pointed out to you?
Hi Lukas
So, the patch led to the initial discussion was probably this:
https://groups.google.com/g/clang-built-linux/c/1C6YoJKBsQQ/m/a8IS1NjGAgAJ
This patch is related to changes made in tools/objtool. So, we probably cannot test this patch for AVOID_L_PREFIX warning.
I'll walk you through the discussion:
Because of the above changes in objtool, [PATCH] x86/entry: use STB_GLOBAL for register restoring thunk was proposed
(https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/qWnKkZWFBgAJ)
It was later decided that the changes can be instead made to .L prefix local symbols here for avoiding the objtool warning:
https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/MtjkJK6LCgAJ
And for standardizing this practice, the suggestion for checkpatch rule was made here:
https://groups.google.com/g/clang-built-linux/c/-drkmLgu-cU/m/4staOlf-CgAJ
However, I have tested the patch on arch/x86/ files and also the patch proposed here fixing for one of those files.
As expected, I got no warnings for the patch fixing the warnings for thunk_64.S file, and some warnings for files where the prefix was used.
Thanks
Aditya
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH 1/1 checkpatch: add warning for avoiding .L prefix symbols in assembly files
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.
--- scripts/checkpatch.pl | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
scripts/checkpatch.pl +22 -0
@@ -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)$/);
Signature · 6 lines
2.17.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 7 lines · 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.
Quoted 7 lines
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 <[email protected]>Signed-off-by Aditya Srivastava <[email protected]>---
Okay, send it out to the initial discussion group (with the standard CCs for mentees and checkpatch) and lkml. Then we see.
Lukas
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees