WARNING: Found insecure W+X mapping in v6.16-rc6 (Bisected)
Subsystems (2)
Hello memory management folks,
I just recently started testing the v6.16-rc6 kernel, and started getting a new warning about insecure W+X mappings:
[ 6.022847] ------------[ cut here ]------------ [ 6.023020] x86/mm: Found insecure W+X mapping at address 0xffffffffc0247000 [ 6.023200] WARNING: CPU: 5 PID: 1 at arch/x86/mm/dump_pagetables.c:246 note_page+0x6ec/0x790 [ 6.023381] CPU: 5 UID: 0 PID: 1 Comm: swapper/0 Tainted: G T 6.16.0-rc6 #1 PREEMPT [ 6.023558] Tainted: [T]=RANDSTRUCT [ 6.023734] Hardware name: Dell Inc. XPS 8930/0T2HR0, BIOS 1.1.31 11/21/2023 [ 6.023913] RIP: 0010:note_page+0x6ec/0x790 [ 6.024094] Code: 85 6c ff ff ff e9 ed fb ff ff 80 3d ae a9 c3 01 00 0f 85 c5 f9 ff ff 48 c7 c7 40 3c 5d 8c c6 05 9a a9 c3 01 01 e8 84 01 02 00 <0f> 0b e9 ab f9 ff ff 48 c7 c6 2d 2e 65 8c 4c 89 e7 e8 2e f7 20 00 [ 6.024290] RSP: 0000:ffffc15fc0033bf0 EFLAGS: 00010246 [ 6.024486] RAX: 0000000000000000 RBX: 0000000000000004 RCX: 0000000000000000 [ 6.024684] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000 [ 6.024884] RBP: ffffffffc0248000 R08: 0000000000000000 R09: 0000000000000000 [ 6.025082] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000 [ 6.025277] R13: 0000000000000000 R14: 0000000000000000 R15: ffffc15fc0033e38 [ 6.025475] FS: 0000000000000000(0000) GS:ffff9b54736ba000(0000) knlGS:0000000000000000 [ 6.025674] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 6.025874] CR2: 0000000000000000 CR3: 00000004ab818001 CR4: 00000000003726f0 [ 6.026076] Call Trace: [ 6.026277] <TASK> [ 6.026474] ptdump_pte_entry+0x30/0x50 [ 6.026678] walk_pte_range_inner+0x9c/0xe0 [ 6.026882] walk_pgd_range+0x5c7/0xc10 [ 6.027082] ? __pfx_ptdump_pte_entry+0x10/0x10 [ 6.027284] walk_page_range_novma+0x9f/0xc0 [ 6.027484] ptdump_walk_pgd+0x4f/0x90 [ 6.027684] ptdump_walk_pgd_level_core+0x1b0/0x1f0 [ 6.027886] ? __pfx_note_page_pte+0x10/0x10 [ 6.028085] ? __pfx_note_page_pmd+0x10/0x10 [ 6.028281] ? __pfx_note_page_pud+0x10/0x10 [ 6.028474] ? __pfx_note_page_p4d+0x10/0x10 [ 6.028666] ? __pfx_note_page_pgd+0x10/0x10 [ 6.028858] ? __pfx_note_page_flush+0x10/0x10 [ 6.029046] ? __pfx_effective_prot_pte+0x10/0x10 [ 6.029233] ? __pfx_effective_prot_pmd+0x10/0x10 [ 6.029418] ? __pfx_effective_prot_pud+0x10/0x10 [ 6.029601] ? __pfx_effective_prot_p4d+0x10/0x10 [ 6.029780] ? __pfx_effective_prot_pgd+0x10/0x10 [ 6.029957] ? __pfx_kernel_init+0x10/0x10 [ 6.030133] kernel_init+0x3d/0x120 [ 6.030306] ret_from_fork+0x71/0xc0 [ 6.030476] ? __pfx_kernel_init+0x10/0x10 [ 6.030642] ret_from_fork_asm+0x1a/0x30 [ 6.030811] </TASK> [ 6.030972] ---[ end trace 0000000000000000 ]--- [ 6.031153] x86/mm: Checked W+X mappings: FAILED, 10 W+X pages found.
I bisected it and git-bisect identified one of three possible commits (I couldn't test these individually because builds at those points fail to boot on my machine):
There are only 'skip'ped commits left to test. The first bad commit could be any of:
0b0cae7119a0ec9449d7261b5e672a5fed765068 47410d839fcda6890cb82828f874f97710982f24 a82b26451de126a5ae130361081986bc459afe9b We cannot bisect more!
I then tried individually reverting each of these and found that if I revert 47410d839fcd ("x86/Kconfig: only enable ROX cache in execmem when STRICT_MODULE_RWX is set") on top of v6.16-rc6 then the warning goes away. I looked at the change a little to understand why it might be causing a problem, and I think it might be because my kernel configuration does not enable loadable module support at all, so in my configuration STRICT_MODULE_RWX isn't enabled and is probably irrelevant given I don't build loadable modules.
Please let me know if there are any additional details I can provide or if I should run additional tests with debugging options enabled, etc.
Cheers,
-- Dan
Hi,
Quoted 8 lines · Dan Moulding
Hello memory management folks,
I just recently started testing the v6.16-rc6 kernel, and started getting a new warning about insecure W+X mappings:
[ 6.022847] ------------[ cut here ]------------ [ 6.023020] x86/mm: Found insecure W+X mapping at address 0xffffffffc0247000 [ 6.023200] WARNING: CPU: 5 PID: 1 at arch/x86/mm/dump_pagetables.c:246 note_page+0x6ec/0x790
...
Quoted 11 lines
[ 6.031153] x86/mm: Checked W+X mappings: FAILED, 10 W+X pages found.I bisected it and git-bisect identified one of three possible commits (I couldn't test these individually because builds at those points fail to boot on my machine):
There are only 'skip'ped commits left to test. The first bad commit could be any of: 0b0cae7119a0ec9449d7261b5e672a5fed765068 47410d839fcda6890cb82828f874f97710982f24 a82b26451de126a5ae130361081986bc459afe9b
I'd say this one is the culprit: a82b26451de1 ("x86/its: explicitly manage permissions for ITS pages")
Quoted 11 lines
We cannot bisect more!
I then tried individually reverting each of these and found that if I revert
47410d839fcd("x86/Kconfig: only enable ROX cache in execmem whenSTRICT_MODULE_RWXis set") on top of v6.16-rc6 then the warning goes away. I looked at the change a little to understand why it might be causing a problem, and I think it might be because my kernel configuration does not enable loadable module support at all, so in my configurationSTRICT_MODULE_RWXisn't enabled and is probably irrelevant given I don't build loadable modules.
Quoted 2 lines
Please let me know if there are any additional details I can provide or if I should run additional tests with debugging options enabled, etc.
Can you please try this patch:
arch/x86/Kconfig +1 -1
@@ -89,7 +89,7 @@ config X86 select ARCH_HAS_DMA_OPS if GART_IOMMU || XEN select ARCH_HAS_EARLY_DEBUG if KGDB select ARCH_HAS_ELF_RANDOMIZE - select ARCH_HAS_EXECMEM_ROX if X86_64 && STRICT_MODULE_RWX + select ARCH_HAS_EXECMEM_ROX if X86_64 && (STRICT_KERNEL_RWX || STRICT_MODULE_RWX) select ARCH_HAS_FAST_MULTIPLIER select ARCH_HAS_FORTIFY_SOURCE select ARCH_HAS_GCOV_PROFILE_ALL
Quoted 3 lines
Cheers,
-- Dan
Signature · 2 lines
Sincerely yours, Mike.
Quoted 7 lines
Can you please try this patch:
arch/x86/Kconfig +0
-0
arch/x86/Kconfig config X86 · line 95
Quoted 8 lines
arch/x86/Kconfig +1
-1@@ -89,7 +89,7 @@ config X86 select ARCH_HAS_DMA_OPS if GART_IOMMU || XEN select ARCH_HAS_EARLY_DEBUG if KGDB select ARCH_HAS_ELF_RANDOMIZE - select ARCH_HAS_EXECMEM_ROX if X86_64 && STRICT_MODULE_RWX + select ARCH_HAS_EXECMEM_ROX if X86_64 && (STRICT_KERNEL_RWX || STRICT_MODULE_RWX) select ARCH_HAS_FAST_MULTIPLIER select ARCH_HAS_FORTIFY_SOURCE select ARCH_HAS_GCOV_PROFILE_ALL
Yep, this patch works. I applied it directly to v6.16-rc6 and the warning goes away.
Thanks!
-- Dan
Quoted 55 lines · Mike Rapoport
Hi,
Quoted 8 lines · Dan Moulding
Hello memory management folks,
I just recently started testing the v6.16-rc6 kernel, and started getting a new warning about insecure W+X mappings:
[ 6.022847] ------------[ cut here ]------------ [ 6.023020] x86/mm: Found insecure W+X mapping at address 0xffffffffc0247000 [ 6.023200] WARNING: CPU: 5 PID: 1 at arch/x86/mm/dump_pagetables.c:246 note_page+0x6ec/0x790...
Quoted 11 lines
[ 6.031153] x86/mm: Checked W+X mappings: FAILED, 10 W+X pages found.I bisected it and git-bisect identified one of three possible commits (I couldn't test these individually because builds at those points fail to boot on my machine):
There are only 'skip'ped commits left to test. The first bad commit could be any of: 0b0cae7119a0ec9449d7261b5e672a5fed765068 47410d839fcda6890cb82828f874f97710982f24 a82b26451de126a5ae130361081986bc459afe9bI'd say this one is the culprit:
a82b26451de1("x86/its: explicitly manage permissions for ITS pages")Quoted 11 lines
We cannot bisect more!
I then tried individually reverting each of these and found that if I revert
47410d839fcd("x86/Kconfig: only enable ROX cache in execmem whenSTRICT_MODULE_RWXis set") on top of v6.16-rc6 then the warning goes away. I looked at the change a little to understand why it might be causing a problem, and I think it might be because my kernel configuration does not enable loadable module support at all, so in my configurationSTRICT_MODULE_RWXisn't enabled and is probably irrelevant given I don't build loadable modules.Quoted 2 lines
Please let me know if there are any additional details I can provide or if I should run additional tests with debugging options enabled, etc.
Can you please try this patch:
arch/x86/Kconfig +1
-1@@ -89,7 +89,7 @@ config X86 select ARCH_HAS_DMA_OPS if GART_IOMMU || XEN select ARCH_HAS_EARLY_DEBUG if KGDB select ARCH_HAS_ELF_RANDOMIZE - select ARCH_HAS_EXECMEM_ROX if X86_64 && STRICT_MODULE_RWX + select ARCH_HAS_EXECMEM_ROX if X86_64 && (STRICT_KERNEL_RWX || STRICT_MODULE_RWX) select ARCH_HAS_FAST_MULTIPLIER select ARCH_HAS_FORTIFY_SOURCE select ARCH_HAS_GCOV_PROFILE_ALL
Quoted 3 lines
Quoted 3 lines
Cheers,
-- Dan
Hello,
I've hit this too when building the kernel with CONFIG_MODULES=n (not because I need them disabled, just while doing random tests)
I suppose "(STRICT_KERNEL_RWX || STRICT_MODULE_RWX)" is redundant as x86 always selects STRICT_KERNEL_RWX.
Since I see that STRICT_MODULE_RWX check was added some time ago to correct an issue with module mapping permissions:47410d839fcd ("x86/Kconfig: only enable ROX cache in execmem when STRICT_MODULE_RWX is set")
then perhaps something like this would make sense:
arch/x86/Kconfig +0 -0
@@ -84,7 +84,7 @@ config X86
select ARCH_HAS_DMA_OPS if GART_IOMMU || XEN select ARCH_HAS_EARLY_DEBUG if KGDB select ARCH_HAS_ELF_RANDOMIZE
- select ARCH_HAS_EXECMEM_ROX if X86_64 && STRICT_MODULE_RWX
+ select ARCH_HAS_EXECMEM_ROX if X86_64 && (STRICT_MODULE_RWX || !MODULES)
select ARCH_HAS_FAST_MULTIPLIER
select ARCH_HAS_FORTIFY_SOURCE
select ARCH_HAS_GCOV_PROFILE_ALL