Orphan section warnings from `.text.asan.module_{c,d}tor` #1432

nathanchance · 2021-07-25T23:37:36Z

After llvm/llvm-project@7b78956, enabling at least CONFIG_KASAN results in tons of orphan section warnings:

$ make -skj"$(nproc)" LLVM=1 LLVM_IAS=1 distclean defconfig

$ scripts/config -e KASAN

$ make -skj"$(nproc)" LLVM=1 LLVM_IAS=1 olddefconfig all
...
ld.lld: warning: arch/x86/kernel/ebda.o:(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor'
ld.lld: warning: arch/x86/kernel/platform-quirks.o:(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor'
ld.lld: warning: init/built-in.a(main.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor'
ld.lld: warning: init/built-in.a(main.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor'
ld.lld: warning: init/built-in.a(version.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor'
ld.lld: warning: init/built-in.a(version.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor'
ld.lld: warning: init/built-in.a(do_mounts.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor'
ld.lld: warning: init/built-in.a(do_mounts.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor'
...

Should these sections be kept or should we discard them? It seems that according to the commit message, prior to this commit, they were not being kept? If we should keep them, should they just go in TEXT_TEXT within include/asm-generic/vmlinux.lds.h?

cc @MaskRay @kees

The text was updated successfully, but these errors were encountered:

MaskRay · 2021-07-26T00:51:21Z

clang -fsanitize=kernel-address emits a no-op function asan.module_ctor. Its text section can just be discarded.

Hmm: I haven't tested why it did not warn previously. Does a kernel tool previously somehow discard .text.asan.module_ctor?

The only clang difference is that the section has the SHF_GNU_RETAIN flag now.

nathanchance · 2021-07-26T21:07:42Z

Hmm: I haven't tested why it did not warn previously. Does a kernel tool previously somehow discard .text.asan.module_ctor?

.text.asan nor module_ctor are referenced in any of the lds files (tested with fd -e lds.h -e lds.S . -x rg -Hn 'asan|module_ctor') so it seems like it has never been handled, which means we should have seen a warning from ld.lld prior to this if the section was there, no? If it is not needed, we can just discard it and I can send a patch for that tomorrow.

nathanchance · 2021-07-28T23:27:59Z

diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index ddb2ff158321..521d5e1516f4 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -1009,11 +1009,13 @@
        defined(CONFIG_CFI_CLANG)
 # ifdef CONFIG_CONSTRUCTORS
 #  define SANITIZER_DISCARDS                                           \
-       *(.eh_frame)
+       *(.text.asan.module_ctor) *(.text.asan.module_dtor)             \
+       *(.text.tsan.module_ctor) *(.eh_frame)
 # else
 #  define SANITIZER_DISCARDS                                           \
-       *(.init_array) *(.init_array.*)                                 \
-       *(.eh_frame)
+       *(.text.asan.module_ctor) *(.text.asan.module_dtor)             \
+       *(.text.tsan.module_ctor) *(.eh_frame)                          \
+       *(.init_array) *(.init_array.*)
 # endif
 #else
 # define SANITIZER_DISCARDS

This is what I have tentatively come up with, I will add a comment that references the LLVM commit once I understand everything. -fsanitize=thread (used with CONFIG_KCSAN) produces .text.tsan.module_ctor sections, are these safe to discard as well?

For what it's worth, I ran llvm-readelf --sections init/main.o before and after that LLVM commit, which produces:

Prior to 7b789562244ee941b7bf2cefeb3fc08a59a01865

There are 73 section headers, starting at offset 0x17c58:

Section Headers:
  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            0000000000000000 000000 000000 00      0   0  0
  [ 1] .strtab           STRTAB          0000000000000000 000040 002054 00      0   0  1
  [ 2] .text             PROGBITS        0000000000000000 0020a0 000e92 00  AX  0   0 16
  [ 3] .rela.text        RELA            0000000000000000 002f38 0012d8 18     65   2  8
  [ 4] __tracepoints_ptrs PROGBITS       0000000000000000 004210 00000c 00   A  0   0  4
  [ 5] .rela__tracepoints_ptrs RELA      0000000000000000 004220 000048 18     65   4  8
  [ 6] .static_call.text PROGBITS        0000000000000000 004268 000015 00  AX  0   0  4
  [ 7] .rela.static_call.text RELA       0000000000000000 004280 000048 18     65   6  8
  [ 8] __ksymtab_strings PROGBITS        0000000000000000 0042c8 000046 01 AMS  0   0  1
  [ 9] ___ksymtab+system_state PROGBITS  0000000000000000 004310 00000c 00   A  0   0  4
  [10] .rela___ksymtab+system_state RELA 0000000000000000 004320 000048 18     65   9  8
  [11] ___ksymtab_gpl+static_key_initialized PROGBITS 0000000000000000 004368 00000c 00   A  0   0  4
  [12] .rela___ksymtab_gpl+static_key_initialized RELA 0000000000000000 004378 000048 18     65  11  8
  [13] ___ksymtab+reset_devices PROGBITS 0000000000000000 0043c0 00000c 00   A  0   0  4
  [14] .rela___ksymtab+reset_devices RELA 0000000000000000 0043d0 000048 18     65  13  8
  [15] ___ksymtab+loops_per_jiffy PROGBITS 0000000000000000 004418 00000c 00   A  0   0  4
  [16] .rela___ksymtab+loops_per_jiffy RELA 0000000000000000 004428 000048 18     65  15  8
  [17] .init.text        PROGBITS        0000000000000000 004470 0013e6 00  AX  0   0  1
  [18] .rela.init.text   RELA            0000000000000000 005858 002988 18     65  17  8
  [19] .ref.text         PROGBITS        0000000000000000 0081e0 00026e 00  AX  0   0 16
  [20] .rela.ref.text    RELA            0000000000000000 008450 0006f0 18     65  19  8
  [21] __bug_table       PROGBITS        0000000000000000 008b40 00003c 00  WA  0   0  1
  [22] .rela__bug_table  RELA            0000000000000000 008b80 0000f0 18     65  21  8
  [23] .discard.reachable PROGBITS       0000000000000000 008c70 000010 00      0   0  1
  [24] .rela.discard.reachable RELA      0000000000000000 008c80 000060 18     65  23  8
  [25] __jump_table      PROGBITS        0000000000000000 008ce0 000070 00  WA  0   0  8
  [26] .rela__jump_table RELA            0000000000000000 008d50 0001f8 18     65  25  8
  [27] .discard.unreachable PROGBITS     0000000000000000 008f48 000004 00      0   0  1
  [28] .rela.discard.unreachable RELA    0000000000000000 008f50 000018 18     65  27  8
  [29] .smp_locks        PROGBITS        0000000000000000 008f68 000004 00   A  0   0  4
  [30] .rela.smp_locks   RELA            0000000000000000 008f70 000018 18     65  29  8
  [31] __tracepoints_strings PROGBITS    0000000000000000 008f90 000030 00   A  0   0 16
  [32] .data             PROGBITS        0000000000000000 008fc0 002180 00  WA  0   0 32
  [33] .rela.data        RELA            0000000000000000 00b140 002880 18     65  32  8
  [34] __tracepoints     PROGBITS        0000000000000000 00d9c0 0000d8 00  WA  0   0  8
  [35] .rela__tracepoints RELA           0000000000000000 00da98 000120 18     65  34  8
  [36] .rodata           PROGBITS        0000000000000000 00dbc0 001588 00   A  0   0 32
  [37] .rela.rodata      RELA            0000000000000000 00f148 000768 18     65  36  8
  [38] .ref.data         PROGBITS        0000000000000000 00f8b0 0000d8 00  WA  0   0  8
  [39] .rela.ref.data    RELA            0000000000000000 00f988 000240 18     65  38  8
  [40] _ftrace_events    PROGBITS        0000000000000000 00fbc8 000018 00  WA  0   0  8
  [41] .rela_ftrace_events RELA          0000000000000000 00fbe0 000048 18     65  40  8
  [42] .data..read_mostly PROGBITS       0000000000000000 00fc28 000006 00  WA  0   0  4
  [43] .discard.addressable PROGBITS     0000000000000000 00fc30 000038 00  WA  0   0  8
  [44] .rela.discard.addressable RELA    0000000000000000 00fc68 0000a8 18     65  43  8
  [45] .bss              NOBITS          0000000000000000 00fd20 000100 00  WA  0   0 32
  [46] .init.rodata      PROGBITS        0000000000000000 00fd20 000070 00   A  0   0  1
  [47] .init.setup       PROGBITS        0000000000000000 00fd90 0000f0 00  WA  0   0  8
  [48] .rela.init.setup  RELA            0000000000000000 00fe80 0001e0 18     65  47  8
  [49] .init.data        PROGBITS        0000000000000000 010060 001038 00  WA  0   0 16
  [50] .rela.init.data   RELA            0000000000000000 011098 000030 18     65  49  8
  [51] __param           PROGBITS        0000000000000000 0110c8 000028 00   A  0   0  8
  [52] .rela__param      RELA            0000000000000000 0110f0 000048 18     65  51  8
  [53] .data..ro_after_init PROGBITS     0000000000000000 011138 000011 00  WA  0   0  8
  [54] .data..percpu     PROGBITS        0000000000000000 01114c 000004 00  WA  0   0  4
  [55] .rodata.str1.1    PROGBITS        0000000000000000 011150 000347 01 AMS  0   0  1
  [56] .init.data        PROGBITS        0000000000000000 0114a0 000088 00   A  0   0 16
  [57] .rela.init.data   RELA            0000000000000000 011528 000198 18     65  56  8
  [58] .init_array.1     INIT_ARRAY      0000000000000000 0116c0 000008 00  WA  0   0  8
  [59] .rela.init_array.1 RELA           0000000000000000 0116c8 000018 18     65  58  8
  [60] .fini_array.1     FINI_ARRAY      0000000000000000 0116e0 000008 00  WA  0   0  8
  [61] .rela.fini_array.1 RELA           0000000000000000 0116e8 000018 18     65  60  8
  [62] .comment          PROGBITS        0000000000000000 011700 000076 01  MS  0   0  1
  [63] .note.GNU-stack   PROGBITS        0000000000000000 011776 000000 00      0   0  1
  [64] .llvm_addrsig     LLVM_ADDRSIG    0000000000000000 011776 00017a 00   E 65   0  1
  [65] .symtab           SYMTAB          0000000000000000 0118f0 002b98 18      1 212  8
  [66] .altinstructions  PROGBITS        0000000000000000 014488 000078 0c   A  0   0  1
  [67] .rela.altinstructions RELA        0000000000000000 014500 0001e0 18   I 65  66  8
  [68] .static_call_sites PROGBITS       0000000000000000 0146e0 000020 08  WA  0   0  1
  [69] .rela.static_call_sites RELA      0000000000000000 014700 0000c0 18   I 65  68  8
  [70] .orc_unwind       PROGBITS        0000000000000000 0147c0 000948 06   A  0   0  1
  [71] .orc_unwind_ip    PROGBITS        0000000000000000 015108 000630 04   A  0   0  1
  [72] .rela.orc_unwind_ip RELA          0000000000000000 015738 002520 18   I 65  71  8
Key to Flags:
  W (write), A (alloc), X (execute), M (merge), S (strings), I (info),
  L (link order), O (extra OS processing required), G (group), T (TLS),
  C (compressed), x (unknown), o (OS specific), E (exclude),
  R (retain), l (large), p (processor specific)

After 7b789562244ee941b7bf2cefeb3fc08a59a01865

There are 77 section headers, starting at offset 0x17ca8:

Section Headers:
  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            0000000000000000 000000 000000 00      0   0  0
  [ 1] .strtab           STRTAB          0000000000000000 000040 00206a 00      0   0  1
  [ 2] .text             PROGBITS        0000000000000000 0020b0 000e5f 00  AX  0   0 16
  [ 3] .rela.text        RELA            0000000000000000 002f10 001278 18     69   2  8
  [ 4] __tracepoints_ptrs PROGBITS       0000000000000000 004188 00000c 00   A  0   0  4
  [ 5] .rela__tracepoints_ptrs RELA      0000000000000000 004198 000048 18     69   4  8
  [ 6] .static_call.text PROGBITS        0000000000000000 0041e0 000015 00  AX  0   0  4
  [ 7] .rela.static_call.text RELA       0000000000000000 0041f8 000048 18     69   6  8
  [ 8] __ksymtab_strings PROGBITS        0000000000000000 004240 000046 01 AMS  0   0  1
  [ 9] ___ksymtab+system_state PROGBITS  0000000000000000 004288 00000c 00   A  0   0  4
  [10] .rela___ksymtab+system_state RELA 0000000000000000 004298 000048 18     69   9  8
  [11] ___ksymtab_gpl+static_key_initialized PROGBITS 0000000000000000 0042e0 00000c 00   A  0   0  4
  [12] .rela___ksymtab_gpl+static_key_initialized RELA 0000000000000000 0042f0 000048 18     69  11  8
  [13] ___ksymtab+reset_devices PROGBITS 0000000000000000 004338 00000c 00   A  0   0  4
  [14] .rela___ksymtab+reset_devices RELA 0000000000000000 004348 000048 18     69  13  8
  [15] ___ksymtab+loops_per_jiffy PROGBITS 0000000000000000 004390 00000c 00   A  0   0  4
  [16] .rela___ksymtab+loops_per_jiffy RELA 0000000000000000 0043a0 000048 18     69  15  8
  [17] .init.text        PROGBITS        0000000000000000 0043e8 0013e6 00  AX  0   0  1
  [18] .rela.init.text   RELA            0000000000000000 0057d0 002988 18     69  17  8
  [19] .ref.text         PROGBITS        0000000000000000 008160 00026e 00  AX  0   0 16
  [20] .rela.ref.text    RELA            0000000000000000 0083d0 0006f0 18     69  19  8
  [21] __bug_table       PROGBITS        0000000000000000 008ac0 00003c 00  WA  0   0  1
  [22] .rela__bug_table  RELA            0000000000000000 008b00 0000f0 18     69  21  8
  [23] .discard.reachable PROGBITS       0000000000000000 008bf0 000010 00      0   0  1
  [24] .rela.discard.reachable RELA      0000000000000000 008c00 000060 18     69  23  8
  [25] __jump_table      PROGBITS        0000000000000000 008c60 000070 00  WA  0   0  8
  [26] .rela__jump_table RELA            0000000000000000 008cd0 0001f8 18     69  25  8
  [27] .discard.unreachable PROGBITS     0000000000000000 008ec8 000004 00      0   0  1
  [28] .rela.discard.unreachable RELA    0000000000000000 008ed0 000018 18     69  27  8
  [29] .smp_locks        PROGBITS        0000000000000000 008ee8 000004 00   A  0   0  4
  [30] .rela.smp_locks   RELA            0000000000000000 008ef0 000018 18     69  29  8
  [31] .text.asan.module_ctor PROGBITS   0000000000000000 008f10 000012 00 AXR  0   0 16
  [32] .rela.text.asan.module_ctor RELA  0000000000000000 008f28 000030 18     69  31  8
  [33] .text.asan.module_dtor PROGBITS   0000000000000000 008f60 000012 00 AXR  0   0 16
  [34] .rela.text.asan.module_dtor RELA  0000000000000000 008f78 000030 18     69  33  8
  [35] __tracepoints_strings PROGBITS    0000000000000000 008fb0 000030 00   A  0   0 16
  [36] .data             PROGBITS        0000000000000000 008fe0 002180 00  WA  0   0 32
  [37] .rela.data        RELA            0000000000000000 00b160 002880 18     69  36  8
  [38] __tracepoints     PROGBITS        0000000000000000 00d9e0 0000d8 00  WA  0   0  8
  [39] .rela__tracepoints RELA           0000000000000000 00dab8 000120 18     69  38  8
  [40] .rodata           PROGBITS        0000000000000000 00dbe0 001588 00   A  0   0 32
  [41] .rela.rodata      RELA            0000000000000000 00f168 000768 18     69  40  8
  [42] .ref.data         PROGBITS        0000000000000000 00f8d0 0000d8 00  WA  0   0  8
  [43] .rela.ref.data    RELA            0000000000000000 00f9a8 000240 18     69  42  8
  [44] _ftrace_events    PROGBITS        0000000000000000 00fbe8 000018 00  WA  0   0  8
  [45] .rela_ftrace_events RELA          0000000000000000 00fc00 000048 18     69  44  8
  [46] .data..read_mostly PROGBITS       0000000000000000 00fc48 000006 00  WA  0   0  4
  [47] .discard.addressable PROGBITS     0000000000000000 00fc50 000038 00  WA  0   0  8
  [48] .rela.discard.addressable RELA    0000000000000000 00fc88 0000a8 18     69  47  8
  [49] .bss              NOBITS          0000000000000000 00fd40 000100 00  WA  0   0 32
  [50] .init.rodata      PROGBITS        0000000000000000 00fd40 000070 00   A  0   0  1
  [51] .init.setup       PROGBITS        0000000000000000 00fdb0 0000f0 00  WA  0   0  8
  [52] .rela.init.setup  RELA            0000000000000000 00fea0 0001e0 18     69  51  8
  [53] .init.data        PROGBITS        0000000000000000 010080 001038 00  WA  0   0 16
  [54] .rela.init.data   RELA            0000000000000000 0110b8 000030 18     69  53  8
  [55] __param           PROGBITS        0000000000000000 0110e8 000028 00   A  0   0  8
  [56] .rela__param      RELA            0000000000000000 011110 000048 18     69  55  8
  [57] .data..ro_after_init PROGBITS     0000000000000000 011158 000011 00  WA  0   0  8
  [58] .data..percpu     PROGBITS        0000000000000000 01116c 000004 00  WA  0   0  4
  [59] .rodata.str1.1    PROGBITS        0000000000000000 011170 000347 01 AMS  0   0  1
  [60] .init.data        PROGBITS        0000000000000000 0114c0 000088 00   A  0   0 16
  [61] .rela.init.data   RELA            0000000000000000 011548 000198 18     69  60  8
  [62] .init_array.1     INIT_ARRAY      0000000000000000 0116e0 000008 00  WA  0   0  8
  [63] .rela.init_array.1 RELA           0000000000000000 0116e8 000018 18     69  62  8
  [64] .fini_array.1     FINI_ARRAY      0000000000000000 011700 000008 00  WA  0   0  8
  [65] .rela.fini_array.1 RELA           0000000000000000 011708 000018 18     69  64  8
  [66] .comment          PROGBITS        0000000000000000 011720 000076 01  MS  0   0  1
  [67] .note.GNU-stack   PROGBITS        0000000000000000 011796 000000 00      0   0  1
  [68] .llvm_addrsig     LLVM_ADDRSIG    0000000000000000 011796 00017a 00   E 69   0  1
  [69] .symtab           SYMTAB          0000000000000000 011910 002bc8 18      1 214  8
  [70] .altinstructions  PROGBITS        0000000000000000 0144d8 000078 0c   A  0   0  1
  [71] .rela.altinstructions RELA        0000000000000000 014550 0001e0 18   I 69  70  8
  [72] .static_call_sites PROGBITS       0000000000000000 014730 000020 08  WA  0   0  1
  [73] .rela.static_call_sites RELA      0000000000000000 014750 0000c0 18   I 69  72  8
  [74] .orc_unwind       PROGBITS        0000000000000000 014810 000948 06   A  0   0  1
  [75] .orc_unwind_ip    PROGBITS        0000000000000000 015158 000630 04   A  0   0  1
  [76] .rela.orc_unwind_ip RELA          0000000000000000 015788 002520 18   I 69  75  8
Key to Flags:
  W (write), A (alloc), X (execute), M (merge), S (strings), I (info),
  L (link order), O (extra OS processing required), G (group), T (TLS),
  C (compressed), x (unknown), o (OS specific), E (exclude),
  R (retain), l (large), p (processor specific)

The .rela. sections are already handled by individual linker scripts.

nickdesaulniers · 2021-07-29T20:15:26Z

are these safe to discard as well?

I would boot test those changes.

nathanchance · 2021-07-30T02:31:12Z

I would boot test those changes.

Good thing that I did because CONFIG_KASAN + CONFIG_KASAN_KUNIT_TEST panicked:

[    4.019913] BUG: kernel NULL pointer dereference, address: 0000000000000000
[    4.020160] #PF: supervisor instruction fetch in kernel mode
[    4.020160] #PF: error_code(0x0010) - not-present page
[    4.020160] PGD 0 P4D 0
[    4.020160] Oops: 0010 [#1] SMP KASAN NOPTI
[    4.020160] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.14.0-rc3+ #1
[    4.020160] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
[    4.020160] RIP: 0010:0x0
[    4.020160] Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.
[    4.020160] RSP: 0018:ffff888001197ed8 EFLAGS: 00010246
[    4.020160] RAX: ffffffff8a6670b7 RBX: ffffffff8a6670b0 RCX: ffffffff8a57f9c2
[    4.020160] RDX: 0000000000000000 RSI: dffffc0000000000 RDI: ffffffff8a6670b0
[    4.020160] RBP: 0000000000000286 R08: dffffc0000000000 R09: fffffbfff133db41
[    4.020160] R10: fffffbfff133db41 R11: 0000000000000000 R12: 0000000000000001
[    4.020160] R13: 0000000000000000 R14: ffffffff8a66be38 R15: ffffffff8a6670b0
[    4.020160] FS:  0000000000000000(0000) GS:ffff888019a00000(0000) knlGS:0000000000000000
[    4.020160] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    4.020160] CR2: ffffffffffffffd6 CR3: 0000000010610000 CR4: 0000000000350ef0
[    4.020160] Call Trace:
[    4.020160]  do_basic_setup+0x3e/0x59
[    4.020160]  kernel_init_freeable+0x165/0x1cd
[    4.020160]  ? rest_init+0xd0/0xd0
[    4.020160]  kernel_init+0x14/0x1a0
[    4.020160]  ? rest_init+0xd0/0xd0
[    4.020160]  ret_from_fork+0x22/0x30
[    4.020160] Modules linked in:
[    4.020160] CR2: 0000000000000000
[    4.020160] ---[ end trace 26d40edcf555a205 ]---
[    4.020160] RIP: 0010:0x0
[    4.020160] Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.
[    4.020160] RSP: 0018:ffff888001197ed8 EFLAGS: 00010246
[    4.020160] RAX: ffffffff8a6670b7 RBX: ffffffff8a6670b0 RCX: ffffffff8a57f9c2
[    4.020160] RDX: 0000000000000000 RSI: dffffc0000000000 RDI: ffffffff8a6670b0
[    4.020160] RBP: 0000000000000286 R08: dffffc0000000000 R09: fffffbfff133db41
[    4.020160] R10: fffffbfff133db41 R11: 0000000000000000 R12: 0000000000000001
[    4.020160] R13: 0000000000000000 R14: ffffffff8a66be38 R15: ffffffff8a6670b0
[    4.020160] FS:  0000000000000000(0000) GS:ffff888019a00000(0000) knlGS:0000000000000000
[    4.020160] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    4.020160] CR2: ffffffffffffffd6 CR3: 0000000010610000 CR4: 0000000000350ef0
[    4.020160] Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000009
[    4.020160] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000009 ]---

Keeping .text.asan.* resolved the issue and all the tests still pass:

[    9.153243] ok 1 - kasan

This is the diff I will send along tomorrow:

diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index 17325416e2de..3b79b1e76556 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -586,6 +586,7 @@
                NOINSTR_TEXT                                            \
                *(.text..refcount)                                      \
                *(.ref.text)                                            \
+               *(.text.asan .text.asan.*)                              \
                TEXT_CFI_JT                                             \
        MEM_KEEP(init.text*)                                            \
        MEM_KEEP(exit.text*)                                            \

nathanchance · 2021-07-30T22:38:58Z

https://lore.kernel.org/r/20210730223815.1382706-1-nathan@kernel.org/

A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org>

A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org

nathanchance · 2021-08-12T18:14:58Z

Merged into mainline, thanks @kees!

https://git.kernel.org/torvalds/c/848378812e40

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Resolve conflict due to lack of cf68fff] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 848378812e40152abe9b9baf58ce2004f76fb988 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Resolve conflict due to lack of cf68fffb66d60] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fff and 266ff2a] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Resolve conflict due to lack of cf68fff] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 848378812e40152abe9b9baf58ce2004f76fb988 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fffb66d60 and 266ff2a8f51f0] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fff and 266ff2a] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fff and 266ff2a] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 8483788 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Resolve conflict due to lack of cf68fff] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

commit 848378812e40152abe9b9baf58ce2004f76fb988 upstream. A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fffb66d60 and 266ff2a8f51f0] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Kneba <abenkenary3@gmail.com>

A recent change in LLVM causes module_{c,d}tor sections to appear when CONFIG_K{A,C}SAN are enabled, which results in orphan section warnings because these are not handled anywhere: ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_ctor) is being placed in '.text.asan.module_ctor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.asan.module_dtor) is being placed in '.text.asan.module_dtor' ld.lld: warning: arch/x86/pci/built-in.a(legacy.o):(.text.tsan.module_ctor) is being placed in '.text.tsan.module_ctor' Fangrui explains: "the function asan.module_ctor has the SHF_GNU_RETAIN flag, so it is in a separate section even with -fno-function-sections (default)". Place them in the TEXT_TEXT section so that these technologies continue to work with the newer compiler versions. All of the KASAN and KCSAN KUnit tests continue to pass after this change. Cc: stable@vger.kernel.org Link: ClangBuiltLinux/linux#1432 Link: llvm/llvm-project@7b78956 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Fangrui Song <maskray@google.com> Acked-by: Marco Elver <elver@google.com> Signed-off-by: Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/20210731023107.1932981-1-nathan@kernel.org [nc: Fix conflicts due to lack of cf68fffb66d60 and 266ff2a8f51f0] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: shygosh <mossynine@gmail.com>