Re: [PATCH v5 0/6] Fix some bugs related to ramp and dax

[Muchun Song <songmuchun@bytedance.com>]

On Thu, Mar 31, 2022 at 11:55 PM Qian Cai <quic_qiancai@quicinc.com> wrote:
>
> On Fri, Mar 18, 2022 at 03:45:23PM +0800, Muchun Song wrote:
> > This series is based on next-20220225.
> >
> > Patch 1-2 fix a cache flush bug, because subsequent patches depend on
> > those on those changes, there are placed in this series.  Patch 3-4
> > are preparation for fixing a dax bug in patch 5.  Patch 6 is code cleanup
> > since the previous patch remove the usage of follow_invalidate_pte().
>
> Reverting this series fixed boot crashes.
>
>  KASAN: null-ptr-deref in range [0x0000000000000018-0x000000000000001f]
>  Mem abort info:
>    ESR = 0x96000004
>    EC = 0x25: DABT (current EL), IL = 32 bits
>    SET = 0, FnV = 0
>    EA = 0, S1PTW = 0
>    FSC = 0x04: level 0 translation fault
>  Data abort info:
>    ISV = 0, ISS = 0x00000004
>    CM = 0, WnR = 0
>  [dfff800000000003] address between user and kernel address ranges
>  Internal error: Oops: 96000004 [#1] PREEMPT SMP
>  Modules linked in: cdc_ether usbnet ipmi_devintf ipmi_msghandler cppc_cpufreq fuse ip_tables x_tables ipv6 btrfs blake2b_generic libcrc32c xor xor_neon raid6_pq zstd_compress dm_mod nouveau crct10dif_ce drm_ttm_helper mlx5_core ttm drm_dp_helper drm_kms_helper nvme mpt3sas nvme_core xhci_pci raid_class drm xhci_pci_renesas
>  CPU: 3 PID: 1707 Comm: systemd-udevd Not tainted 5.17.0-next-20220331-00004-g2d550916a6b9 #51
>  pstate: 104000c9 (nzcV daIF +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>  pc : __lock_acquire
>  lr : lock_acquire.part.0
>  sp : ffff800030a16fd0
>  x29: ffff800030a16fd0 x28: ffffdd876c4e9f90 x27: 0000000000000018
>  x26: 0000000000000000 x25: 0000000000000018 x24: 0000000000000000
>  x23: ffff08022beacf00 x22: ffffdd8772507660 x21: 0000000000000000
>  x20: 0000000000000000 x19: 0000000000000000 x18: ffffdd8772417d2c
>  x17: ffffdd876c5bc2e0 x16: 1fffe100457d5b06 x15: 0000000000000094
>  x14: 000000000000f1f1 x13: 00000000f3f3f3f3 x12: ffff08022beacf08
>  x11: 1ffffbb0ee482fa5 x10: ffffdd8772417d28 x9 : 0000000000000000
>  x8 : 0000000000000003 x7 : ffffdd876c4e9f90 x6 : 0000000000000000
>  x5 : 0000000000000000 x4 : 0000000000000001 x3 : 0000000000000000
>  x2 : 0000000000000000 x1 : 0000000000000003 x0 : dfff800000000000
>  Call trace:
>   __lock_acquire
>   lock_acquire.part.0
>   lock_acquire
>   _raw_spin_lock
>   page_vma_mapped_walk
>   try_to_migrate_one
>   rmap_walk_anon
>   try_to_migrate
>   __unmap_and_move
>   unmap_and_move
>   migrate_pages
>   migrate_misplaced_page
>   do_huge_pmd_numa_page
>   __handle_mm_fault
>   handle_mm_fault
>   do_translation_fault
>   do_mem_abort
>   el0_da
>   el0t_64_sync_handler
>   el0t_64_sync
>  Code: d65f03c0 d343ff61 d2d00000 f2fbffe0 (38e06820)

Hi,

I have found the root cause. It is because the implementation of
pmd_leaf() on arm64 is wrong.  It didn't consider the PROT_NONE
mapped PMD, which does not match the expectation of pmd_leaf().
I'll send a fixed patch for arm64 like the following.

diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 94e147e5456c..09eaae46a19b 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -535,7 +535,7 @@ extern pgprot_t phys_mem_access_prot(struct file
*file, unsigned long pfn,
                                 PMD_TYPE_TABLE)
 #define pmd_sect(pmd)          ((pmd_val(pmd) & PMD_TYPE_MASK) == \
                                 PMD_TYPE_SECT)
-#define pmd_leaf(pmd)          pmd_sect(pmd)
+#define pmd_leaf(pmd)          (pmd_present(pmd) && !(pmd_val(pmd) &
PMD_TABLE_BIT))
 #define pmd_bad(pmd)           (!pmd_table(pmd))

 #define pmd_leaf_size(pmd)     (pmd_cont(pmd) ? CONT_PMD_SIZE : PMD_SIZE)

Thanks.