From b3d09d06e052e1d754645acea4e4d1e96f81c934 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 18 Feb 2025 14:59:19 -0500 Subject: arm64: scripts/sorttable: Implement sorting mcount_loc at boot for arm64 The mcount_loc section holds the addresses of the functions that get patched by ftrace when enabling function callbacks. It can contain tens of thousands of entries. These addresses must be sorted. If they are not sorted at compile time, they are sorted at boot. Sorting at boot does take some time and does have a small impact on boot performance. x86 and arm32 have the addresses in the mcount_loc section of the ELF file. But for arm64, the section just contains zeros. The .rela.dyn Elf_Rela section holds the addresses and they get patched at boot during the relocation phase. In order to sort these addresses, the Elf_Rela needs to be updated instead of the location in the binary that holds the mcount_loc section. Have the sorttable code, allocate an array to hold the functions, load the addresses from the Elf_Rela entries, sort them, then put them back in order into the Elf_rela entries so that they will be sorted at boot up without having to sort them during boot up. Cc: bpf Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Peter Zijlstra Cc: Linus Torvalds Cc: Masahiro Yamada Cc: Nathan Chancellor Cc: Nicolas Schier Cc: Zheng Yejian Cc: Martin Kelly Cc: Christophe Leroy Cc: Josh Poimboeuf Cc: Heiko Carstens Cc: Will Deacon Cc: Vasily Gorbik Cc: Alexander Gordeev Link: https://lore.kernel.org/20250218200022.373319428@goodmis.org Acked-by: Catalin Marinas Signed-off-by: Steven Rostedt (Google) --- scripts/sorttable.c | 185 +++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 182 insertions(+), 3 deletions(-) (limited to 'scripts') diff --git a/scripts/sorttable.c b/scripts/sorttable.c index 9f41575afd7a..4a34c275123e 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -28,6 +28,7 @@ #include #include #include +#include #include #include #include @@ -79,10 +80,16 @@ typedef union { Elf64_Sym e64; } Elf_Sym; +typedef union { + Elf32_Rela e32; + Elf64_Rela e64; +} Elf_Rela; + static uint32_t (*r)(const uint32_t *); static uint16_t (*r2)(const uint16_t *); static uint64_t (*r8)(const uint64_t *); static void (*w)(uint32_t, uint32_t *); +static void (*w8)(uint64_t, uint64_t *); typedef void (*table_sort_t)(char *, int); static struct elf_funcs { @@ -102,6 +109,10 @@ static struct elf_funcs { uint32_t (*sym_name)(Elf_Sym *sym); uint64_t (*sym_value)(Elf_Sym *sym); uint16_t (*sym_shndx)(Elf_Sym *sym); + uint64_t (*rela_offset)(Elf_Rela *rela); + uint64_t (*rela_info)(Elf_Rela *rela); + uint64_t (*rela_addend)(Elf_Rela *rela); + void (*rela_write_addend)(Elf_Rela *rela, uint64_t val); } e; static uint64_t ehdr64_shoff(Elf_Ehdr *ehdr) @@ -262,6 +273,38 @@ SYM_ADDR(value) SYM_WORD(name) SYM_HALF(shndx) +#define __maybe_unused __attribute__((__unused__)) + +#define RELA_ADDR(fn_name) \ +static uint64_t rela64_##fn_name(Elf_Rela *rela) \ +{ \ + return r8((uint64_t *)&rela->e64.r_##fn_name); \ +} \ + \ +static uint64_t rela32_##fn_name(Elf_Rela *rela) \ +{ \ + return r((uint32_t *)&rela->e32.r_##fn_name); \ +} \ + \ +static uint64_t __maybe_unused rela_##fn_name(Elf_Rela *rela) \ +{ \ + return e.rela_##fn_name(rela); \ +} + +RELA_ADDR(offset) +RELA_ADDR(info) +RELA_ADDR(addend) + +static void rela64_write_addend(Elf_Rela *rela, uint64_t val) +{ + w8(val, (uint64_t *)&rela->e64.r_addend); +} + +static void rela32_write_addend(Elf_Rela *rela, uint64_t val) +{ + w(val, (uint32_t *)&rela->e32.r_addend); +} + /* * Get the whole file as a programming convenience in order to avoid * malloc+lseek+read+free of many pieces. If successful, then mmap @@ -341,6 +384,16 @@ static void wle(uint32_t val, uint32_t *x) put_unaligned_le32(val, x); } +static void w8be(uint64_t val, uint64_t *x) +{ + put_unaligned_be64(val, x); +} + +static void w8le(uint64_t val, uint64_t *x) +{ + put_unaligned_le64(val, x); +} + /* * Move reserved section indices SHN_LORESERVE..SHN_HIRESERVE out of * the way to -256..-1, to avoid conflicting with real section @@ -398,13 +451,12 @@ static inline void *get_index(void *start, int entsize, int index) static int extable_ent_size; static int long_size; +#define ERRSTR_MAXSZ 256 #ifdef UNWINDER_ORC_ENABLED /* ORC unwinder only support X86_64 */ #include -#define ERRSTR_MAXSZ 256 - static char g_err[ERRSTR_MAXSZ]; static int *g_orc_ip_table; static struct orc_entry *g_orc_table; @@ -499,7 +551,19 @@ static void *sort_orctable(void *arg) #endif #ifdef MCOUNT_SORT_ENABLED + +/* Only used for sorting mcount table */ +static void rela_write_addend(Elf_Rela *rela, uint64_t val) +{ + e.rela_write_addend(rela, val); +} + static pthread_t mcount_sort_thread; +static bool sort_reloc; + +static long rela_type; + +static char m_err[ERRSTR_MAXSZ]; struct elf_mcount_loc { Elf_Ehdr *ehdr; @@ -508,6 +572,103 @@ struct elf_mcount_loc { uint64_t stop_mcount_loc; }; +/* Sort the relocations not the address itself */ +static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) +{ + Elf_Shdr *shdr_start; + Elf_Rela *rel; + unsigned int shnum; + unsigned int count; + int shentsize; + void *vals; + void *ptr; + + shdr_start = (Elf_Shdr *)((char *)ehdr + ehdr_shoff(ehdr)); + shentsize = ehdr_shentsize(ehdr); + + vals = malloc(long_size * size); + if (!vals) { + snprintf(m_err, ERRSTR_MAXSZ, "Failed to allocate sort array"); + pthread_exit(m_err); + return NULL; + } + + ptr = vals; + + shnum = ehdr_shnum(ehdr); + if (shnum == SHN_UNDEF) + shnum = shdr_size(shdr_start); + + for (int i = 0; i < shnum; i++) { + Elf_Shdr *shdr = get_index(shdr_start, shentsize, i); + void *end; + + if (shdr_type(shdr) != SHT_RELA) + continue; + + rel = (void *)ehdr + shdr_offset(shdr); + end = (void *)rel + shdr_size(shdr); + + for (; (void *)rel < end; rel = (void *)rel + shdr_entsize(shdr)) { + uint64_t offset = rela_offset(rel); + + if (offset >= start_loc && offset < start_loc + size) { + if (ptr + long_size > vals + size) { + free(vals); + snprintf(m_err, ERRSTR_MAXSZ, + "Too many relocations"); + pthread_exit(m_err); + return NULL; + } + + /* Make sure this has the correct type */ + if (rela_info(rel) != rela_type) { + free(vals); + snprintf(m_err, ERRSTR_MAXSZ, + "rela has type %lx but expected %lx\n", + (long)rela_info(rel), rela_type); + pthread_exit(m_err); + return NULL; + } + + if (long_size == 4) + *(uint32_t *)ptr = rela_addend(rel); + else + *(uint64_t *)ptr = rela_addend(rel); + ptr += long_size; + } + } + } + count = ptr - vals; + qsort(vals, count / long_size, long_size, compare_extable); + + ptr = vals; + for (int i = 0; i < shnum; i++) { + Elf_Shdr *shdr = get_index(shdr_start, shentsize, i); + void *end; + + if (shdr_type(shdr) != SHT_RELA) + continue; + + rel = (void *)ehdr + shdr_offset(shdr); + end = (void *)rel + shdr_size(shdr); + + for (; (void *)rel < end; rel = (void *)rel + shdr_entsize(shdr)) { + uint64_t offset = rela_offset(rel); + + if (offset >= start_loc && offset < start_loc + size) { + if (long_size == 4) + rela_write_addend(rel, *(uint32_t *)ptr); + else + rela_write_addend(rel, *(uint64_t *)ptr); + ptr += long_size; + } + } + } + free(vals); + return NULL; +} + /* Sort the addresses stored between __start_mcount_loc to __stop_mcount_loc in vmlinux */ static void *sort_mcount_loc(void *arg) { @@ -517,6 +678,9 @@ static void *sort_mcount_loc(void *arg) uint64_t count = emloc->stop_mcount_loc - emloc->start_mcount_loc; unsigned char *start_loc = (void *)emloc->ehdr + offset; + if (sort_reloc) + return sort_relocs(emloc->ehdr, emloc->start_mcount_loc, count); + qsort(start_loc, count/long_size, long_size, compare_extable); return NULL; } @@ -866,12 +1030,14 @@ static int do_file(char const *const fname, void *addr) r2 = r2le; r8 = r8le; w = wle; + w8 = w8le; break; case ELFDATA2MSB: r = rbe; r2 = r2be; r8 = r8be; w = wbe; + w8 = w8be; break; default: fprintf(stderr, "unrecognized ELF data encoding %d: %s\n", @@ -887,8 +1053,13 @@ static int do_file(char const *const fname, void *addr) } switch (r2(&ehdr->e32.e_machine)) { - case EM_386: case EM_AARCH64: +#ifdef MCOUNT_SORT_ENABLED + sort_reloc = true; + rela_type = 0x403; +#endif + /* fallthrough */ + case EM_386: case EM_LOONGARCH: case EM_RISCV: case EM_S390: @@ -932,6 +1103,10 @@ static int do_file(char const *const fname, void *addr) .sym_name = sym32_name, .sym_value = sym32_value, .sym_shndx = sym32_shndx, + .rela_offset = rela32_offset, + .rela_info = rela32_info, + .rela_addend = rela32_addend, + .rela_write_addend = rela32_write_addend, }; e = efuncs; @@ -965,6 +1140,10 @@ static int do_file(char const *const fname, void *addr) .sym_name = sym64_name, .sym_value = sym64_value, .sym_shndx = sym64_shndx, + .rela_offset = rela64_offset, + .rela_info = rela64_info, + .rela_addend = rela64_addend, + .rela_write_addend = rela64_write_addend, }; e = efuncs; -- cgit v1.2.3 From a0265659322540d656727b9e132edfb6f06b6c1a Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 18 Feb 2025 14:59:20 -0500 Subject: scripts/sorttable: Have mcount rela sort use direct values The mcount_loc sorting for when the values are stored in the Elf_Rela entries uses the compare_extable() function to do the compares in the qsort(). That function does handle byte swapping if the machine being compiled for is a different endian than the host machine. But the sort_relocs() function sorts an array that pulled in the values from the Elf_Rela section and has already done the swapping. Create two new compare functions that will sort the direct values. One will sort 32 bit values and the other will sort the 64 bit value. One of these will be assigned to a compare_values function pointer and that will be used for sorting the Elf_Rela mcount values. Cc: bpf Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Peter Zijlstra Cc: Linus Torvalds Cc: Masahiro Yamada Cc: Nathan Chancellor Cc: Nicolas Schier Cc: Zheng Yejian Cc: Martin Kelly Cc: Christophe Leroy Cc: Josh Poimboeuf Cc: Heiko Carstens Cc: Catalin Marinas Cc: Will Deacon Cc: Vasily Gorbik Cc: Alexander Gordeev Link: https://lore.kernel.org/20250218200022.538888594@goodmis.org Signed-off-by: Steven Rostedt (Google) --- scripts/sorttable.c | 26 +++++++++++++++++++++++++- 1 file changed, 25 insertions(+), 1 deletion(-) (limited to 'scripts') diff --git a/scripts/sorttable.c b/scripts/sorttable.c index 4a34c275123e..f62a91d8af0a 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -552,6 +552,28 @@ static void *sort_orctable(void *arg) #ifdef MCOUNT_SORT_ENABLED +static int compare_values_64(const void *a, const void *b) +{ + uint64_t av = *(uint64_t *)a; + uint64_t bv = *(uint64_t *)b; + + if (av < bv) + return -1; + return av > bv; +} + +static int compare_values_32(const void *a, const void *b) +{ + uint32_t av = *(uint32_t *)a; + uint32_t bv = *(uint32_t *)b; + + if (av < bv) + return -1; + return av > bv; +} + +static int (*compare_values)(const void *a, const void *b); + /* Only used for sorting mcount table */ static void rela_write_addend(Elf_Rela *rela, uint64_t val) { @@ -583,6 +605,8 @@ static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) void *vals; void *ptr; + compare_values = long_size == 4 ? compare_values_32 : compare_values_64; + shdr_start = (Elf_Shdr *)((char *)ehdr + ehdr_shoff(ehdr)); shentsize = ehdr_shentsize(ehdr); @@ -640,7 +664,7 @@ static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) } } count = ptr - vals; - qsort(vals, count / long_size, long_size, compare_extable); + qsort(vals, count / long_size, long_size, compare_values); ptr = vals; for (int i = 0; i < shnum; i++) { -- cgit v1.2.3 From 5fb964f5ba53afda0e2b6dbc00b8205461ffe04a Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 18 Feb 2025 14:59:21 -0500 Subject: scripts/sorttable: Always use an array for the mcount_loc sorting The sorting of the mcount_loc section is done directly to the section for x86 and arm32 but it uses a separate array for arm64 as arm64 has the values for the mcount_loc stored in the rela sections of the vmlinux ELF file. In order to use the same code to remove weak functions, always use a separate array to do the sorting. This requires splitting up the filling of the array into one function and the placing the contents of the array back into the rela sections or into the mcount_loc section into a separate file. Cc: bpf Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Peter Zijlstra Cc: Linus Torvalds Cc: Masahiro Yamada Cc: Nathan Chancellor Cc: Nicolas Schier Cc: Zheng Yejian Cc: Martin Kelly Cc: Christophe Leroy Cc: Josh Poimboeuf Cc: Heiko Carstens Cc: Catalin Marinas Cc: Will Deacon Cc: Vasily Gorbik Cc: Alexander Gordeev Link: https://lore.kernel.org/20250218200022.710676551@goodmis.org Signed-off-by: Steven Rostedt (Google) --- scripts/sorttable.c | 122 ++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 90 insertions(+), 32 deletions(-) (limited to 'scripts') diff --git a/scripts/sorttable.c b/scripts/sorttable.c index f62a91d8af0a..ec02a2852efb 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -594,31 +594,19 @@ struct elf_mcount_loc { uint64_t stop_mcount_loc; }; -/* Sort the relocations not the address itself */ -static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) +/* Fill the array with the content of the relocs */ +static int fill_relocs(void *ptr, uint64_t size, Elf_Ehdr *ehdr, uint64_t start_loc) { Elf_Shdr *shdr_start; Elf_Rela *rel; unsigned int shnum; - unsigned int count; + unsigned int count = 0; int shentsize; - void *vals; - void *ptr; - - compare_values = long_size == 4 ? compare_values_32 : compare_values_64; + void *array_end = ptr + size; shdr_start = (Elf_Shdr *)((char *)ehdr + ehdr_shoff(ehdr)); shentsize = ehdr_shentsize(ehdr); - vals = malloc(long_size * size); - if (!vals) { - snprintf(m_err, ERRSTR_MAXSZ, "Failed to allocate sort array"); - pthread_exit(m_err); - return NULL; - } - - ptr = vals; - shnum = ehdr_shnum(ehdr); if (shnum == SHN_UNDEF) shnum = shdr_size(shdr_start); @@ -637,22 +625,18 @@ static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) uint64_t offset = rela_offset(rel); if (offset >= start_loc && offset < start_loc + size) { - if (ptr + long_size > vals + size) { - free(vals); + if (ptr + long_size > array_end) { snprintf(m_err, ERRSTR_MAXSZ, "Too many relocations"); - pthread_exit(m_err); - return NULL; + return -1; } /* Make sure this has the correct type */ if (rela_info(rel) != rela_type) { - free(vals); snprintf(m_err, ERRSTR_MAXSZ, "rela has type %lx but expected %lx\n", (long)rela_info(rel), rela_type); - pthread_exit(m_err); - return NULL; + return -1; } if (long_size == 4) @@ -660,13 +644,28 @@ static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) else *(uint64_t *)ptr = rela_addend(rel); ptr += long_size; + count++; } } } - count = ptr - vals; - qsort(vals, count / long_size, long_size, compare_values); + return count; +} + +/* Put the sorted vals back into the relocation elements */ +static void replace_relocs(void *ptr, uint64_t size, Elf_Ehdr *ehdr, uint64_t start_loc) +{ + Elf_Shdr *shdr_start; + Elf_Rela *rel; + unsigned int shnum; + int shentsize; + + shdr_start = (Elf_Shdr *)((char *)ehdr + ehdr_shoff(ehdr)); + shentsize = ehdr_shentsize(ehdr); + + shnum = ehdr_shnum(ehdr); + if (shnum == SHN_UNDEF) + shnum = shdr_size(shdr_start); - ptr = vals; for (int i = 0; i < shnum; i++) { Elf_Shdr *shdr = get_index(shdr_start, shentsize, i); void *end; @@ -689,8 +688,32 @@ static void *sort_relocs(Elf_Ehdr *ehdr, uint64_t start_loc, uint64_t size) } } } - free(vals); - return NULL; +} + +static int fill_addrs(void *ptr, uint64_t size, void *addrs) +{ + void *end = ptr + size; + int count = 0; + + for (; ptr < end; ptr += long_size, addrs += long_size, count++) { + if (long_size == 4) + *(uint32_t *)ptr = r(addrs); + else + *(uint64_t *)ptr = r8(addrs); + } + return count; +} + +static void replace_addrs(void *ptr, uint64_t size, void *addrs) +{ + void *end = ptr + size; + + for (; ptr < end; ptr += long_size, addrs += long_size) { + if (long_size == 4) + w(*(uint32_t *)ptr, addrs); + else + w8(*(uint64_t *)ptr, addrs); + } } /* Sort the addresses stored between __start_mcount_loc to __stop_mcount_loc in vmlinux */ @@ -699,14 +722,49 @@ static void *sort_mcount_loc(void *arg) struct elf_mcount_loc *emloc = (struct elf_mcount_loc *)arg; uint64_t offset = emloc->start_mcount_loc - shdr_addr(emloc->init_data_sec) + shdr_offset(emloc->init_data_sec); - uint64_t count = emloc->stop_mcount_loc - emloc->start_mcount_loc; + uint64_t size = emloc->stop_mcount_loc - emloc->start_mcount_loc; unsigned char *start_loc = (void *)emloc->ehdr + offset; + Elf_Ehdr *ehdr = emloc->ehdr; + void *e_msg = NULL; + void *vals; + int count; + + vals = malloc(long_size * size); + if (!vals) { + snprintf(m_err, ERRSTR_MAXSZ, "Failed to allocate sort array"); + pthread_exit(m_err); + } if (sort_reloc) - return sort_relocs(emloc->ehdr, emloc->start_mcount_loc, count); + count = fill_relocs(vals, size, ehdr, emloc->start_mcount_loc); + else + count = fill_addrs(vals, size, start_loc); + + if (count < 0) { + e_msg = m_err; + goto out; + } + + if (count != size / long_size) { + snprintf(m_err, ERRSTR_MAXSZ, "Expected %u mcount elements but found %u\n", + (int)(size / long_size), count); + e_msg = m_err; + goto out; + } + + compare_values = long_size == 4 ? compare_values_32 : compare_values_64; + + qsort(vals, count, long_size, compare_values); + + if (sort_reloc) + replace_relocs(vals, size, ehdr, emloc->start_mcount_loc); + else + replace_addrs(vals, size, start_loc); + +out: + free(vals); - qsort(start_loc, count/long_size, long_size, compare_extable); - return NULL; + pthread_exit(e_msg); } /* Get the address of __start_mcount_loc and __stop_mcount_loc in System.map */ -- cgit v1.2.3 From ef378c3b8233855497a414b9d67bf22592c928a4 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 18 Feb 2025 14:59:22 -0500 Subject: scripts/sorttable: Zero out weak functions in mcount_loc table When a function is annotated as "weak" and is overridden, the code is not removed. If it is traced, the fentry/mcount location in the weak function will be referenced by the "__mcount_loc" section. This will then be added to the available_filter_functions list. Since only the address of the functions are listed, to find the name to show, a search of kallsyms is used. Since kallsyms will return the function by simply finding the function that the address is after but before the next function, an address of a weak function will show up as the function before it. This is because kallsyms does not save names of weak functions. This has caused issues in the past, as now the traced weak function will be listed in available_filter_functions with the name of the function before it. At best, this will cause the previous function's name to be listed twice. At worse, if the previous function was marked notrace, it will now show up as a function that can be traced. Note that it only shows up that it can be traced but will not be if enabled, which causes confusion. https://lore.kernel.org/all/20220412094923.0abe90955e5db486b7bca279@kernel.org/ The commit b39181f7c6907 ("ftrace: Add FTRACE_MCOUNT_MAX_OFFSET to avoid adding weak function") was a workaround to this by checking the function address before printing its name. If the address was too far from the function given by the name then instead of printing the name it would print: __ftrace_invalid_address___ The real issue is that these invalid addresses are listed in the ftrace table look up which available_filter_functions is derived from. A place holder must be listed in that file because set_ftrace_filter may take a series of indexes into that file instead of names to be able to do O(1) lookups to enable filtering (many tools use this method). Even if kallsyms saved the size of the function, it does not remove the need of having these place holders. The real solution is to not add a weak function into the ftrace table in the first place. To solve this, the sorttable.c code that sorts the mcount regions during the build is modified to take a "nm -S vmlinux" input, sort it, and any function listed in the mcount_loc section that is not within a boundary of the function list given by nm is considered a weak function and is zeroed out. Note, this does not mean they will remain zero when booting as KASLR will still shift those addresses. To handle this, the entries in the mcount_loc section will be ignored if they are zero or match the kaslr_offset() value. Before: ~# grep __ftrace_invalid_address___ /sys/kernel/tracing/available_filter_functions | wc -l 551 After: ~# grep __ftrace_invalid_address___ /sys/kernel/tracing/available_filter_functions | wc -l 0 Cc: bpf Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Peter Zijlstra Cc: Linus Torvalds Cc: Masahiro Yamada Cc: Nathan Chancellor Cc: Nicolas Schier Cc: Zheng Yejian Cc: Martin Kelly Cc: Christophe Leroy Cc: Josh Poimboeuf Cc: Heiko Carstens Cc: Catalin Marinas Cc: Will Deacon Cc: Vasily Gorbik Cc: Alexander Gordeev Link: https://lore.kernel.org/20250218200022.883095980@goodmis.org Signed-off-by: Steven Rostedt (Google) --- scripts/link-vmlinux.sh | 4 +- scripts/sorttable.c | 128 +++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 129 insertions(+), 3 deletions(-) (limited to 'scripts') diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh index 56a077d204cf..59b07fe6fd00 100755 --- a/scripts/link-vmlinux.sh +++ b/scripts/link-vmlinux.sh @@ -177,12 +177,14 @@ mksysmap() sorttable() { - ${objtree}/scripts/sorttable ${1} + ${NM} -S ${1} > .tmp_vmlinux.nm-sort + ${objtree}/scripts/sorttable -s .tmp_vmlinux.nm-sort ${1} } cleanup() { rm -f .btf.* + rm -f .tmp_vmlinux.nm-sort rm -f System.map rm -f vmlinux rm -f vmlinux.map diff --git a/scripts/sorttable.c b/scripts/sorttable.c index ec02a2852efb..23c7e0e6c024 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -580,6 +580,98 @@ static void rela_write_addend(Elf_Rela *rela, uint64_t val) e.rela_write_addend(rela, val); } +struct func_info { + uint64_t addr; + uint64_t size; +}; + +/* List of functions created by: nm -S vmlinux */ +static struct func_info *function_list; +static int function_list_size; + +/* Allocate functions in 1k blocks */ +#define FUNC_BLK_SIZE 1024 +#define FUNC_BLK_MASK (FUNC_BLK_SIZE - 1) + +static int add_field(uint64_t addr, uint64_t size) +{ + struct func_info *fi; + int fsize = function_list_size; + + if (!(fsize & FUNC_BLK_MASK)) { + fsize += FUNC_BLK_SIZE; + fi = realloc(function_list, fsize * sizeof(struct func_info)); + if (!fi) + return -1; + function_list = fi; + } + fi = &function_list[function_list_size++]; + fi->addr = addr; + fi->size = size; + return 0; +} + +/* Only return match if the address lies inside the function size */ +static int cmp_func_addr(const void *K, const void *A) +{ + uint64_t key = *(const uint64_t *)K; + const struct func_info *a = A; + + if (key < a->addr) + return -1; + return key >= a->addr + a->size; +} + +/* Find the function in function list that is bounded by the function size */ +static int find_func(uint64_t key) +{ + return bsearch(&key, function_list, function_list_size, + sizeof(struct func_info), cmp_func_addr) != NULL; +} + +static int cmp_funcs(const void *A, const void *B) +{ + const struct func_info *a = A; + const struct func_info *b = B; + + if (a->addr < b->addr) + return -1; + return a->addr > b->addr; +} + +static int parse_symbols(const char *fname) +{ + FILE *fp; + char addr_str[20]; /* Only need 17, but round up to next int size */ + char size_str[20]; + char type; + + fp = fopen(fname, "r"); + if (!fp) { + perror(fname); + return -1; + } + + while (fscanf(fp, "%16s %16s %c %*s\n", addr_str, size_str, &type) == 3) { + uint64_t addr; + uint64_t size; + + /* Only care about functions */ + if (type != 't' && type != 'T' && type != 'W') + continue; + + addr = strtoull(addr_str, NULL, 16); + size = strtoull(size_str, NULL, 16); + if (add_field(addr, size) < 0) + return -1; + } + fclose(fp); + + qsort(function_list, function_list_size, sizeof(struct func_info), cmp_funcs); + + return 0; +} + static pthread_t mcount_sort_thread; static bool sort_reloc; @@ -752,6 +844,21 @@ static void *sort_mcount_loc(void *arg) goto out; } + /* zero out any locations not found by function list */ + if (function_list_size) { + for (void *ptr = vals; ptr < vals + size; ptr += long_size) { + uint64_t key; + + key = long_size == 4 ? r((uint32_t *)ptr) : r8((uint64_t *)ptr); + if (!find_func(key)) { + if (long_size == 4) + *(uint32_t *)ptr = 0; + else + *(uint64_t *)ptr = 0; + } + } + } + compare_values = long_size == 4 ? compare_values_32 : compare_values_64; qsort(vals, count, long_size, compare_values); @@ -801,6 +908,8 @@ static void get_mcount_loc(struct elf_mcount_loc *emloc, Elf_Shdr *symtab_sec, return; } } +#else /* MCOUNT_SORT_ENABLED */ +static inline int parse_symbols(const char *fname) { return 0; } #endif static int do_sort(Elf_Ehdr *ehdr, @@ -1256,14 +1365,29 @@ int main(int argc, char *argv[]) int i, n_error = 0; /* gcc-4.3.0 false positive complaint */ size_t size = 0; void *addr = NULL; + int c; + + while ((c = getopt(argc, argv, "s:")) >= 0) { + switch (c) { + case 's': + if (parse_symbols(optarg) < 0) { + fprintf(stderr, "Could not parse %s\n", optarg); + return -1; + } + break; + default: + fprintf(stderr, "usage: sorttable [-s nm-file] vmlinux...\n"); + return 0; + } + } - if (argc < 2) { + if ((argc - optind) < 1) { fprintf(stderr, "usage: sorttable vmlinux...\n"); return 0; } /* Process each file in turn, allowing deep failure. */ - for (i = 1; i < argc; i++) { + for (i = optind; i < argc; i++) { addr = mmap_file(argv[i], &size); if (!addr) { ++n_error; -- cgit v1.2.3 From 46514b3c2c17c67cefe84b0c1a59e0aaf6093131 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 25 Feb 2025 13:20:07 -0500 Subject: scripts/sorttable: Use normal sort if theres no relocs in the mcount section When ARM 64 is compiled with gcc, the mcount_loc section will be filled with zeros and the addresses will be located in the Elf_Rela sections. To sort the mcount_loc section, the addresses from the Elf_Rela need to be placed into an array and that is sorted. But when ARM 64 is compiled with clang, it does it the same way as other architectures and leaves the addresses as is in the mcount_loc section. To handle both cases, ARM 64 will first try to sort the Elf_Rela section, and if it doesn't find any functions, it will then fall back to the sorting of the addresses in the mcount_loc section itself. Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Masahiro Yamada Cc: Catalin Marinas Cc: Will Deacon Cc: Mark Brown Link: https://lore.kernel.org/20250225182054.648398403@goodmis.org Fixes: b3d09d06e052 ("arm64: scripts/sorttable: Implement sorting mcount_loc at boot for arm64") Reported-by: "Arnd Bergmann" Tested-by: Nathan Chancellor Closes: https://lore.kernel.org/all/893cd8f1-8585-4d25-bf0f-4197bf872465@app.fastmail.com/ Signed-off-by: Steven Rostedt (Google) --- scripts/sorttable.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'scripts') diff --git a/scripts/sorttable.c b/scripts/sorttable.c index 23c7e0e6c024..07ad8116bc8d 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -827,9 +827,14 @@ static void *sort_mcount_loc(void *arg) pthread_exit(m_err); } - if (sort_reloc) + if (sort_reloc) { count = fill_relocs(vals, size, ehdr, emloc->start_mcount_loc); - else + /* gcc may use relocs to save the addresses, but clang does not. */ + if (!count) { + count = fill_addrs(vals, size, start_loc); + sort_reloc = 0; + } + } else count = fill_addrs(vals, size, start_loc); if (count < 0) { -- cgit v1.2.3 From dc208c69c033d3caba0509da1ae065d2b5ff165f Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 25 Feb 2025 13:20:08 -0500 Subject: scripts/sorttable: Allow matches to functions before function entry ARM 64 uses -fpatchable-function-entry=4,2 which adds padding before the function and the addresses in the mcount_loc point there instead of the function entry that is returned by nm. In order to find a function from nm to make sure it's not an unused weak function, the entries in the mcount_loc section needs to match the entries from nm. Since it can be an instruction before the entry, add a before_func variable that ARM 64 can set to 8, and if the mcount_loc entry is within 8 bytes of the nm function entry, then it will be considered a match. Cc: Masami Hiramatsu Cc: Mark Rutland Cc: Mathieu Desnoyers Cc: Andrew Morton Cc: Masahiro Yamada Cc: Catalin Marinas Cc: Will Deacon Cc: "Arnd Bergmann" Cc: Mark Brown Link: https://lore.kernel.org/20250225182054.815536219@goodmis.org Fixes: ef378c3b82338 ("scripts/sorttable: Zero out weak functions in mcount_loc table") Tested-by: Nathan Chancellor Signed-off-by: Steven Rostedt (Google) --- scripts/sorttable.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'scripts') diff --git a/scripts/sorttable.c b/scripts/sorttable.c index 07ad8116bc8d..7b4b3714b1af 100644 --- a/scripts/sorttable.c +++ b/scripts/sorttable.c @@ -611,13 +611,16 @@ static int add_field(uint64_t addr, uint64_t size) return 0; } +/* Used for when mcount/fentry is before the function entry */ +static int before_func; + /* Only return match if the address lies inside the function size */ static int cmp_func_addr(const void *K, const void *A) { uint64_t key = *(const uint64_t *)K; const struct func_info *a = A; - if (key < a->addr) + if (key + before_func < a->addr) return -1; return key >= a->addr + a->size; } @@ -1253,6 +1256,8 @@ static int do_file(char const *const fname, void *addr) #ifdef MCOUNT_SORT_ENABLED sort_reloc = true; rela_type = 0x403; + /* arm64 uses patchable function entry placing before function */ + before_func = 8; #endif /* fallthrough */ case EM_386: -- cgit v1.2.3