Skip to contentSkip to conversation

kallsyms: Fix kallsyms_selftest failure

PATCH Applied Yonghong Song 6 participants 9 messages Aug 25, 2023 – Sep 21, 2023 Live patching

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 kallsyms: Fix kallsyms_selftest failure

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

  [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
  [    1.877901] kallsyms_selftest: abort

It appears that commit 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions") caused the failure. Commit 8cc32a9bbf29 changed cleanup_symbol_name() based on ".llvm." instead of '.' where ".llvm." is appended to a before-lto-optimization local symbol name. We need to propagate such knowledge in kallsyms_selftest.c as well.

Further more, compare_symbol_name() in kallsyms.c needs change as well. In scripts/kallsyms.c, kallsyms_names and kallsyms_seqs_of_names are used to record symbol names themselves and index to symbol names respectively. For example:
kallsyms_names:

    ...
    __amd_smn_rw._entry       <== seq 1000
    __amd_smn_rw._entry.5     <== seq 1001
    __amd_smn_rw.llvm.<hash>  <== seq 1002
    ...

kallsyms_seqs_of_names are sorted based on cleanup_symbol_name() through, so the order in kallsyms_seqs_of_names actually has

  index 1000:   seq 1002   <== __amd_smn_rw.llvm.<hash> (actual symbol comparison using '__amd_smn_rw')
  index 1001:   seq 1000   <== __amd_smn_rw._entry
  index 1002:   seq 1001   <== __amd_smn_rw._entry.5

Let us say at a particular point, at index 1000, symbol '__amd_smn_rw.llvm.<hash>' is comparing to '__amd_smn_rw._entry' where '__amd_smn_rw._entry' is the one to search e.g., with function kallsyms_on_each_match_symbol(). The current implementation will find out '__amd_smn_rw._entry' is less than '__amd_smn_rw.llvm.<hash>' and then continue to search e.g., index 999 and never found a match although the actual index 1001 is a match.

To fix this issue, let us do cleanup_symbol_name() first and then do comparison. In the above case, comparing '__amd_smn_rw' vs '__amd_smn_rw._entry' and '__amd_smn_rw._entry' being greater than '__amd_smn_rw', the next comparison will be > index 1000 and eventually index 1001 will be hit an a match is found.

For any symbols not having '.llvm.' substr, there is no functionality change for compare_symbol_name().

Fixes 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions")
---
 kernel/kallsyms.c          | 17 +++++++----------
 kernel/kallsyms_selftest.c | 23 +----------------------
 2 files changed, 8 insertions(+), 32 deletions(-)
kernel/kallsyms.c +7 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-	int ret;
-
-	ret = strcmp(name, namebuf);
-	if (!ret)
-		return ret;
-
-	if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-		return 0;
-
-	return ret;
+	/* The kallsyms_seqs_of_names is sorted based on names after
+	 * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+	 * To ensure correct bisection in kallsyms_lookup_names(), do
+	 * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+	 */
+	cleanup_symbol_name(namebuf);
+	return strcmp(name, namebuf);
 }
 
 static unsigned int get_symbol_seq(int index)
kernel/kallsyms_selftest.c +1 -22
index a2e3745d15c4..e05ddc33a752 100644
@@ -196,7 +196,7 @@ static bool match_cleanup_name(const char *s, const char *name)
 	if (!IS_ENABLED(CONFIG_LTO_CLANG))
 		return false;
 
-	p = strchr(s, '.');
+	p = strstr(s, ".llvm.");
 	if (!p)
 		return false;
 
@@ -344,27 +344,6 @@ static int test_kallsyms_basic_function(void)
 			goto failed;
 		}
 
-		/*
-		 * The first '.' may be the initial letter, in which case the
-		 * entire symbol name will be truncated to an empty string in
-		 * cleanup_symbol_name(). Do not test these symbols.
-		 *
-		 * For example:
-		 * cat /proc/kallsyms | awk '{print $3}' | grep -E "^\." | head
-		 * .E_read_words
-		 * .E_leading_bytes
-		 * .E_trailing_bytes
-		 * .E_write_words
-		 * .E_copy
-		 * .str.292.llvm.12122243386960820698
-		 * .str.24.llvm.12122243386960820698
-		 * .str.29.llvm.12122243386960820698
-		 * .str.75.llvm.12122243386960820698
-		 * .str.99.llvm.12122243386960820698
-		 */
-		if (IS_ENABLED(CONFIG_LTO_CLANG) && !namebuf[0])
-			continue;
-
 		lookup_addr = kallsyms_lookup_name(namebuf);
 
 		memset(stat, 0, sizeof(*stat));
Signature · 1 line
2.34.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 47 lines · Yonghong Song

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

 [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
 [    1.877901] kallsyms_selftest: abort

It appears that commit 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions") caused the failure. Commit 8cc32a9bbf29 changed cleanup_symbol_name() based on ".llvm." instead of '.' where ".llvm." is appended to a before-lto-optimization local symbol name. We need to propagate such knowledge in kallsyms_selftest.c as well.

Further more, compare_symbol_name() in kallsyms.c needs change as well. In scripts/kallsyms.c, kallsyms_names and kallsyms_seqs_of_names are used to record symbol names themselves and index to symbol names respectively. For example:
kallsyms_names:

   ...
   __amd_smn_rw._entry       <== seq 1000
   __amd_smn_rw._entry.5     <== seq 1001
   __amd_smn_rw.llvm.<hash>  <== seq 1002
   ...

kallsyms_seqs_of_names are sorted based on cleanup_symbol_name() through, so the order in kallsyms_seqs_of_names actually has

 index 1000:   seq 1002   <== __amd_smn_rw.llvm.<hash> (actual symbol comparison using '__amd_smn_rw')
 index 1001:   seq 1000   <== __amd_smn_rw._entry
 index 1002:   seq 1001   <== __amd_smn_rw._entry.5

Let us say at a particular point, at index 1000, symbol '__amd_smn_rw.llvm.<hash>' is comparing to '__amd_smn_rw._entry' where '__amd_smn_rw._entry' is the one to search e.g., with function kallsyms_on_each_match_symbol(). The current implementation will find out '__amd_smn_rw._entry' is less than '__amd_smn_rw.llvm.<hash>' and then continue to search e.g., index 999 and never found a match although the actual index 1001 is a match.

To fix this issue, let us do cleanup_symbol_name() first and then do comparison. In the above case, comparing '__amd_smn_rw' vs '__amd_smn_rw._entry' and '__amd_smn_rw._entry' being greater than '__amd_smn_rw', the next comparison will be > index 1000 and eventually index 1001 will be hit an a match is found.

Quoted 7 lines

For any symbols not having '.llvm.' substr, there is no functionality change for compare_symbol_name().

Fixes 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions")
Reviewed-by Zhen Lei on v1 1/1
#2 lore raw
In reply to Yonghong Song
Quoted 38 lines · Yonghong Song

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

  [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
  [    1.877901] kallsyms_selftest: abort

It appears that commit 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions") caused the failure. Commit 8cc32a9bbf29 changed cleanup_symbol_name() based on ".llvm." instead of '.' where ".llvm." is appended to a before-lto-optimization local symbol name. We need to propagate such knowledge in kallsyms_selftest.c as well.

Further more, compare_symbol_name() in kallsyms.c needs change as well. In scripts/kallsyms.c, kallsyms_names and kallsyms_seqs_of_names are used to record symbol names themselves and index to symbol names respectively. For example:
kallsyms_names:

    ...
    __amd_smn_rw._entry       <== seq 1000
    __amd_smn_rw._entry.5     <== seq 1001
    __amd_smn_rw.llvm.<hash>  <== seq 1002
    ...

kallsyms_seqs_of_names are sorted based on cleanup_symbol_name() through, so the order in kallsyms_seqs_of_names actually has

  index 1000:   seq 1002   <== __amd_smn_rw.llvm.<hash> (actual symbol comparison using '__amd_smn_rw')
  index 1001:   seq 1000   <== __amd_smn_rw._entry
  index 1002:   seq 1001   <== __amd_smn_rw._entry.5

Let us say at a particular point, at index 1000, symbol '__amd_smn_rw.llvm.<hash>' is comparing to '__amd_smn_rw._entry' where '__amd_smn_rw._entry' is the one to search e.g., with function kallsyms_on_each_match_symbol(). The current implementation will find out '__amd_smn_rw._entry' is less than '__amd_smn_rw.llvm.<hash>' and then continue to search e.g., index 999 and never found a match although the actual index 1001 is a match.

Quoted 7 lines

To fix this issue, let us do cleanup_symbol_name() first and then do comparison. In the above case, comparing '__amd_smn_rw' vs '__amd_smn_rw._entry' and '__amd_smn_rw._entry' being greater than '__amd_smn_rw', the next comparison will be > index 1000 and eventually index 1001 will be hit an a match is found.

For any symbols not having '.llvm.' substr, there is no functionality change for compare_symbol_name().

Quoted 80 lines
Fixes 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions")
---
 kernel/kallsyms.c          | 17 +++++++----------
 kernel/kallsyms_selftest.c | 23 +----------------------
 2 files changed, 8 insertions(+), 32 deletions(-)
kernel/kallsyms.c +7 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-	int ret;
-
-	ret = strcmp(name, namebuf);
-	if (!ret)
-		return ret;
-
-	if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-		return 0;
-
-	return ret;
+	/* The kallsyms_seqs_of_names is sorted based on names after
+	 * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+	 * To ensure correct bisection in kallsyms_lookup_names(), do
+	 * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+	 */
+	cleanup_symbol_name(namebuf);
+	return strcmp(name, namebuf);
 }
 
 static unsigned int get_symbol_seq(int index)
kernel/kallsyms_selftest.c +1 -22
index a2e3745d15c4..e05ddc33a752 100644
@@ -196,7 +196,7 @@ static bool match_cleanup_name(const char *s, const char *name)
 	if (!IS_ENABLED(CONFIG_LTO_CLANG))
 		return false;
 
-	p = strchr(s, '.');
+	p = strstr(s, ".llvm.");
 	if (!p)
 		return false;
 
@@ -344,27 +344,6 @@ static int test_kallsyms_basic_function(void)
 			goto failed;
 		}
 
-		/*
-		 * The first '.' may be the initial letter, in which case the
-		 * entire symbol name will be truncated to an empty string in
-		 * cleanup_symbol_name(). Do not test these symbols.
-		 *
-		 * For example:
-		 * cat /proc/kallsyms | awk '{print $3}' | grep -E "^\." | head
-		 * .E_read_words
-		 * .E_leading_bytes
-		 * .E_trailing_bytes
-		 * .E_write_words
-		 * .E_copy
-		 * .str.292.llvm.12122243386960820698
-		 * .str.24.llvm.12122243386960820698
-		 * .str.29.llvm.12122243386960820698
-		 * .str.75.llvm.12122243386960820698
-		 * .str.99.llvm.12122243386960820698
-		 */
-		if (IS_ENABLED(CONFIG_LTO_CLANG) && !namebuf[0])
-			continue;
-
 		lookup_addr = kallsyms_lookup_name(namebuf);
 
 		memset(stat, 0, sizeof(*stat));
Signature · 2 lines
Regards,
  Zhen Lei
maintainer
#3 lore raw

Adding live-patching list into Cc.

Quoted 77 lines · Leizhen (ThunderTown)
Quoted 38 lines · Yonghong Song

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

  [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
  [    1.877901] kallsyms_selftest: abort

It appears that commit 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions") caused the failure. Commit 8cc32a9bbf29 changed cleanup_symbol_name() based on ".llvm." instead of '.' where ".llvm." is appended to a before-lto-optimization local symbol name. We need to propagate such knowledge in kallsyms_selftest.c as well.

Further more, compare_symbol_name() in kallsyms.c needs change as well. In scripts/kallsyms.c, kallsyms_names and kallsyms_seqs_of_names are used to record symbol names themselves and index to symbol names respectively. For example:
kallsyms_names:

    ...
    __amd_smn_rw._entry       <== seq 1000
    __amd_smn_rw._entry.5     <== seq 1001
    __amd_smn_rw.llvm.<hash>  <== seq 1002
    ...

kallsyms_seqs_of_names are sorted based on cleanup_symbol_name() through, so the order in kallsyms_seqs_of_names actually has

  index 1000:   seq 1002   <== __amd_smn_rw.llvm.<hash> (actual symbol comparison using '__amd_smn_rw')
  index 1001:   seq 1000   <== __amd_smn_rw._entry
  index 1002:   seq 1001   <== __amd_smn_rw._entry.5

Let us say at a particular point, at index 1000, symbol '__amd_smn_rw.llvm.<hash>' is comparing to '__amd_smn_rw._entry' where '__amd_smn_rw._entry' is the one to search e.g., with function kallsyms_on_each_match_symbol(). The current implementation will find out '__amd_smn_rw._entry' is less than '__amd_smn_rw.llvm.<hash>' and then continue to search e.g., index 999 and never found a match although the actual index 1001 is a match.

Quoted 7 lines

To fix this issue, let us do cleanup_symbol_name() first and then do comparison. In the above case, comparing '__amd_smn_rw' vs '__amd_smn_rw._entry' and '__amd_smn_rw._entry' being greater than '__amd_smn_rw', the next comparison will be > index 1000 and eventually index 1001 will be hit an a match is found.

For any symbols not having '.llvm.' substr, there is no functionality change for compare_symbol_name().

Quoted 28 lines
Fixes 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions")
---
 kernel/kallsyms.c          | 17 +++++++----------
 kernel/kallsyms_selftest.c | 23 +----------------------
 2 files changed, 8 insertions(+), 32 deletions(-)
kernel/kallsyms.c +0 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-	int ret;
-
-	ret = strcmp(name, namebuf);
-	if (!ret)
-		return ret;
-
-	if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-		return 0;
-
-	return ret;
Quoted 8 lines

kernel/kallsyms.c static bool cleanup_symbol_name(char *s) · line 198

Quoted 8 lines
kernel/kallsyms.c +7 -0
+	/* The kallsyms_seqs_of_names is sorted based on names after
+	 * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+	 * To ensure correct bisection in kallsyms_lookup_names(), do
+	 * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+	 */
+	cleanup_symbol_name(namebuf);
+	return strcmp(name, namebuf);
 }

Hmm, I think that this is not the right fix.

The problem is that compare_symbol_name() does not longer allow to match the full name of the extra .llwm. symbols.

I think that the problem is that the problem is that the symbols are sorted using cleanup_symbol_name(). They should be sorted by using the full name.

Note that the original compare_symbol_name() returned return value when comparing the non-stripped name. It will work correctly when the non-stripped names are sorted.

I believe that the correct fix is:

scripts/kallsyms.c +0 -23
index 653b92f6d4c8..da1f8ae68999 100644
@@ -339,25 +339,6 @@ static int symbol_absolute(const struct sym_entry *s)
 	return s->percpu_absolute;
 }
 
-static void cleanup_symbol_name(char *s)
-{
-	char *p;
-
-	/*
-	 * ASCII[.]   = 2e
-	 * ASCII[0-9] = 30,39
-	 * ASCII[A-Z] = 41,5a
-	 * ASCII[_]   = 5f
-	 * ASCII[a-z] = 61,7a
-	 *
-	 * As above, replacing the first '.' in ".llvm." with '\0' does not
-	 * affect the main sorting, but it helps us with subsorting.
-	 */
-	p = strstr(s, ".llvm.");
-	if (p)
-		*p = '\0';
-}
-
 static int compare_names(const void *a, const void *b)
 {
 	int ret;
@@ -533,10 +514,6 @@ static void write_src(void)
 		printf("\n");
 	}
 
-	if (lto_clang)
-		for (i = 0; i < table_cnt; i++)
-			cleanup_symbol_name((char *)table[i]->sym);
-
 	sort_symbols_by_name();
 	output_label("kallsyms_seqs_of_names");
 	for (i = 0; i < table_cnt; i++)

Unfortunately, I could not check it easily because I do not have any experience with building kernel with C-lang.

Anyway, what do you think, please?

Best Regards,
Petr

#4 lore raw
Quoted 147 lines · Petr Mladek

Adding live-patching list into Cc.

Quoted 77 lines · Leizhen (ThunderTown)
Quoted 38 lines · Yonghong Song

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

  [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
  [    1.877901] kallsyms_selftest: abort

It appears that commit 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions") caused the failure. Commit 8cc32a9bbf29 changed cleanup_symbol_name() based on ".llvm." instead of '.' where ".llvm." is appended to a before-lto-optimization local symbol name. We need to propagate such knowledge in kallsyms_selftest.c as well.

Further more, compare_symbol_name() in kallsyms.c needs change as well. In scripts/kallsyms.c, kallsyms_names and kallsyms_seqs_of_names are used to record symbol names themselves and index to symbol names respectively. For example:
kallsyms_names:

    ...
    __amd_smn_rw._entry       <== seq 1000
    __amd_smn_rw._entry.5     <== seq 1001
    __amd_smn_rw.llvm.<hash>  <== seq 1002
    ...

kallsyms_seqs_of_names are sorted based on cleanup_symbol_name() through, so the order in kallsyms_seqs_of_names actually has

  index 1000:   seq 1002   <== __amd_smn_rw.llvm.<hash> (actual symbol comparison using '__amd_smn_rw')
  index 1001:   seq 1000   <== __amd_smn_rw._entry
  index 1002:   seq 1001   <== __amd_smn_rw._entry.5

Let us say at a particular point, at index 1000, symbol '__amd_smn_rw.llvm.<hash>' is comparing to '__amd_smn_rw._entry' where '__amd_smn_rw._entry' is the one to search e.g., with function kallsyms_on_each_match_symbol(). The current implementation will find out '__amd_smn_rw._entry' is less than '__amd_smn_rw.llvm.<hash>' and then continue to search e.g., index 999 and never found a match although the actual index 1001 is a match.

Quoted 7 lines

To fix this issue, let us do cleanup_symbol_name() first and then do comparison. In the above case, comparing '__amd_smn_rw' vs '__amd_smn_rw._entry' and '__amd_smn_rw._entry' being greater than '__amd_smn_rw', the next comparison will be > index 1000 and eventually index 1001 will be hit an a match is found.

For any symbols not having '.llvm.' substr, there is no functionality change for compare_symbol_name().

Quoted 28 lines
Fixes 8cc32a9bbf29 ("kallsyms: strip LTO-only suffixes from promoted global functions")
---
 kernel/kallsyms.c          | 17 +++++++----------
 kernel/kallsyms_selftest.c | 23 +----------------------
 2 files changed, 8 insertions(+), 32 deletions(-)
kernel/kallsyms.c +0 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-	int ret;
-
-	ret = strcmp(name, namebuf);
-	if (!ret)
-		return ret;
-
-	if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-		return 0;
-
-	return ret;
Quoted 8 lines

kernel/kallsyms.c static bool cleanup_symbol_name(char *s) · line 198

Quoted 8 lines
kernel/kallsyms.c +7 -0
+	/* The kallsyms_seqs_of_names is sorted based on names after
+	 * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+	 * To ensure correct bisection in kallsyms_lookup_names(), do
+	 * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+	 */
+	cleanup_symbol_name(namebuf);
+	return strcmp(name, namebuf);
 }

Hmm, I think that this is not the right fix.

The problem is that compare_symbol_name() does not longer allow to match the full name of the extra .llwm. symbols.

I think that the problem is that the problem is that the symbols are sorted using cleanup_symbol_name(). They should be sorted by using the full name.

Note that the original compare_symbol_name() returned return value when comparing the non-stripped name. It will work correctly when the non-stripped names are sorted.

I believe that the correct fix is:

scripts/kallsyms.c +0 -23
index 653b92f6d4c8..da1f8ae68999 100644
@@ -339,25 +339,6 @@ static int symbol_absolute(const struct sym_entry *s)
 	return s->percpu_absolute;
 }
 
-static void cleanup_symbol_name(char *s)
-{
-	char *p;
-
-	/*
-	 * ASCII[.]   = 2e
-	 * ASCII[0-9] = 30,39
-	 * ASCII[A-Z] = 41,5a
-	 * ASCII[_]   = 5f
-	 * ASCII[a-z] = 61,7a
-	 *
-	 * As above, replacing the first '.' in ".llvm." with '\0' does not
-	 * affect the main sorting, but it helps us with subsorting.
-	 */
-	p = strstr(s, ".llvm.");
-	if (p)
-		*p = '\0';
-}
-
 static int compare_names(const void *a, const void *b)
 {
 	int ret;
@@ -533,10 +514,6 @@ static void write_src(void)
 		printf("\n");
 	}
 
-	if (lto_clang)
-		for (i = 0; i < table_cnt; i++)
-			cleanup_symbol_name((char *)table[i]->sym);
-
 	sort_symbols_by_name();
 	output_label("kallsyms_seqs_of_names");
 	for (i = 0; i < table_cnt; i++)
Quoted 2 lines

Unfortunately, I could not check it easily because I do not have any experience with building kernel with C-lang.

make CC=clang distclean defconfig
make CC=clang -j64 2>err.txt

make LLVM=1 LLVM_IAS=1 distclean defconfig
scripts/config -e LTO_CLANG_THIN
scripts/config -e KALLSYMS_SELFTEST
make LLVM=1 LLVM_IAS=1 -j64

Quoted 2 lines

Anyway, what do you think, please?

I'm busy these days. There's a lot of things before the holiday. I'll think about it over the holidays.

Quoted 5 lines

Best Regards,
Petr
.

Signature · 2 lines
Regards,
  Zhen Lei
#5 lore raw
In reply to Yonghong Song
Quoted 15 lines · Yonghong Song
kernel/kallsyms.c +0 -6
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-       int ret;
-
-       ret = strcmp(name, namebuf);
-       if (!ret)
-               return ret;
-

kernel/kallsyms.c static bool cleanup_symbol_name(char *s) · line 196

Quoted 10 lines
kernel/kallsyms.c +6 -4
-       if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-               return 0;
-
-       return ret;
+       /* The kallsyms_seqs_of_names is sorted based on names after
+        * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+        * To ensure correct bisection in kallsyms_lookup_names(), do
+        * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+        */
+       cleanup_symbol_name(namebuf);

Hi Yonghong,
Thanks for your work on this patch.
So if this change is removing the last place where the return value of cleanup_symbol_name is checked, then perhaps this commit should additionally change the function signature of cleanup_symbol_name to have void return type.

Signature · 2 lines
Thanks,
~Nick Desaulniers
#6 lore raw
Quoted 25 lines · Nick Desaulniers
Quoted 24 lines · Yonghong Song
kernel/kallsyms.c +5 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
 static int compare_symbol_name(const char *name, char *namebuf)
 {
-       int ret;
-
-       ret = strcmp(name, namebuf);
-       if (!ret)
-               return ret;
-
-       if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-               return 0;
-
-       return ret;
+       /* The kallsyms_seqs_of_names is sorted based on names after
+        * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+        * To ensure correct bisection in kallsyms_lookup_names(), do
+        * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+        */
Quoted 8 lines

kernel/kallsyms.c static bool cleanup_symbol_name(char *s) · line 196

Quoted 1 line
kernel/kallsyms.c +1 -0
+       cleanup_symbol_name(namebuf);

Hi Yonghong,
Thanks for your work on this patch.
So if this change is removing the last place where the return value of cleanup_symbol_name is checked, then perhaps this commit should additionally change the function signature of cleanup_symbol_name to have void return type.

I've landed this in -next as-is just because I want to make sure the bug gets fixed ASAP, so if this gets adjusted, I can just include that change on top.

Signature · 1 line
Kees Cook
#7 lore raw
Quoted 35 lines · Kees Cook
Quoted 25 lines · Nick Desaulniers
Quoted 24 lines · Yonghong Song
kernel/kallsyms.c +5 -10
index 016d997131d4..e12d26c10dba 100644
@@ -188,16 +188,13 @@ static bool cleanup_symbol_name(char *s)
 
  static int compare_symbol_name(const char *name, char *namebuf)
  {
-       int ret;
-
-       ret = strcmp(name, namebuf);
-       if (!ret)
-               return ret;
-
-       if (cleanup_symbol_name(namebuf) && !strcmp(name, namebuf))
-               return 0;
-
-       return ret;
+       /* The kallsyms_seqs_of_names is sorted based on names after
+        * cleanup_symbol_name() (see scripts/kallsyms.c) if clang lto is enabled.
+        * To ensure correct bisection in kallsyms_lookup_names(), do
+        * cleanup_symbol_name(namebuf) before comparing name and namebuf.
+        */
Quoted 8 lines

kernel/kallsyms.c static bool cleanup_symbol_name(char *s) · line 196

Quoted 1 line
kernel/kallsyms.c +1 -0
+       cleanup_symbol_name(namebuf);

Hi Yonghong,
Thanks for your work on this patch.
So if this change is removing the last place where the return value of cleanup_symbol_name is checked, then perhaps this commit should additionally change the function signature of cleanup_symbol_name to have void return type.

Quoted 3 lines

I've landed this in -next as-is just because I want to make sure the bug gets fixed ASAP, so if this gets adjusted, I can just include that change on top.

Thanks, Kees! I can provide a followup soon.

Applied
#8 lore raw
In reply to Yonghong Song
Quoted 8 lines · Yonghong Song

Kernel test robot reported a kallsyms_test failure when clang lto is enabled (thin or full) and CONFIG_KALLSYMS_SELFTEST is also enabled. I can reproduce in my local environment with the following error message with thin lto:

  [    1.877897] kallsyms_selftest: Test for 1750th symbol failed: (tsc_cs_mark_unstable) addr=ffffffff81038090
  [    1.877901] kallsyms_selftest: abort

[...]

Applied to for-next/hardening, thanks!

[1/1] kallsyms: Fix kallsyms_selftest failure

      https://git.kernel.org/kees/c/33f0467fe069

Take care,

Signature · 1 line
Kees Cook