mce: fix set_mce_nospec to always unmap the whole page

The set_memory_uc() approach doesn't work well in all cases.
As Dan pointed out when "The VMM unmapped the bad page from
guest physical space and passed the machine check to the guest."
"The guest gets virtual #MC on an access to that page. When
the guest tries to do set_memory_uc() and instructs cpa_flush()
to do clean caches that results in taking another fault / exception
perhaps because the VMM unmapped the page from the guest."

Since the driver has special knowledge to handle NP or UC,
mark the poisoned page with NP and let driver handle it when
it comes down to repair.

Please refer to discussions here for more details.
https://lore.kernel.org/all/CAPcyv4hrXPb1tASBZUg-GgdVs0OOFKXMXLiHmktg_kFi7YBMyQ@mail.gmail.com/

Now since poisoned page is marked as not-present, in order to
avoid writing to a not-present page and trigger kernel Oops,
also fix pmem_do_write().

Fixes: 284ce4011b ("x86/memory_failure: Introduce {set, clear}_mce_nospec()")
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Jane Chu <jane.chu@oracle.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/165272615484.103830.2563950688772226611.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This commit is contained in:
Jane Chu
2022-05-16 11:38:10 -07:00
committed by Dan Williams
parent b3fdf9398a
commit 5898b43af9
4 changed files with 24 additions and 41 deletions

View File

@@ -43,10 +43,10 @@ static inline bool can_set_direct_map(void)
#endif /* CONFIG_ARCH_HAS_SET_DIRECT_MAP */
#ifdef CONFIG_X86_64
int set_mce_nospec(unsigned long pfn, bool unmap);
int set_mce_nospec(unsigned long pfn);
int clear_mce_nospec(unsigned long pfn);
#else
static inline int set_mce_nospec(unsigned long pfn, bool unmap)
static inline int set_mce_nospec(unsigned long pfn)
{
return 0;
}