kernel-packages team mailing list archive

["Joshua R. Poulson" <jopoulso@xxxxxxxxxxxxx>]

I'm expecting an upstream fix for this problem soon, based on this
discussion (lkml):

>From Dexuan Cui:
"d1cd12108346: x86, pageattr: Prevent overflow in slow_virt_to_phys() for X86_PAE"
was unintentionally removed by the recent
"34437e67a672: x86/mm: Fix slow_virt_to_phys() to handle large PAT bit".

And, the variable 'phys_addr' was defined as "unsigned long" by mistake -- it should
be "phys_addr_t".

As a result, Hyper-V network driver in 32-PAE Linux guest can't work
again.

Fixes: "commmit 34437e67a672: x86/mm: Fix slow_virt_to_phys() to handle large PAT bit"
Signed-off-by: Dexuan Cui <decui@xxxxxxxxxxxxx>
Cc: Toshi Kani <toshi.kani@xxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: K. Y. Srinivasan <kys@xxxxxxxxxxxxx>
Cc: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
Cc: gregkh@xxxxxxxxxxxxxxxxxxx
Cc: linux-mm@xxxxxxxxx
Cc: olaf@xxxxxxxxx
Cc: apw@xxxxxxxxxxxxx
Cc: jasowang@xxxxxxxxxx
Cc: stable@xxxxxxxxxxxxxxx
---
 arch/x86/mm/pageattr.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 2440814..9cf96d8 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -419,24 +419,30 @@ pmd_t *lookup_pmd_address(unsigned long address)
 phys_addr_t slow_virt_to_phys(void *__virt_addr)
 {
        unsigned long virt_addr = (unsigned long)__virt_addr;
-       unsigned long phys_addr, offset;
+       phys_addr_t phys_addr;
+       unsigned long offset;
        enum pg_level level;
        pte_t *pte;

        pte = lookup_address(virt_addr, &level);
        BUG_ON(!pte);

+       /*
+        * pXX_pfn() returns unsigned long, which must be cast to phys_addr_t
+        * before being left-shifted PAGE_SHIFT bits -- this trick is to
+        * make 32-PAE kernel work correctly.
+        */
        switch (level) {
        case PG_LEVEL_1G:
-               phys_addr = pud_pfn(*(pud_t *)pte) << PAGE_SHIFT;
+               phys_addr = (phys_addr_t)pud_pfn(*(pud_t *)pte) << PAGE_SHIFT;
                offset = virt_addr & ~PUD_PAGE_MASK;
                break;
        case PG_LEVEL_2M:
-               phys_addr = pmd_pfn(*(pmd_t *)pte) << PAGE_SHIFT;
+               phys_addr = (phys_addr_t)pmd_pfn(*(pmd_t *)pte) << PAGE_SHIFT;
                offset = virt_addr & ~PMD_PAGE_MASK;
                break;
        default:
-               phys_addr = pte_pfn(*pte) << PAGE_SHIFT;
+               phys_addr = (phys_addr_t)pte_pfn(*pte) << PAGE_SHIFT;
                offset = virt_addr & ~PAGE_MASK;
        }

>From Toshi Kani:
On Thu, 2016-02-25 at 01:58 -0800, Dexuan Cui wrote:
> "d1cd12108346: x86, pageattr: Prevent overflow in slow_virt_to_phys() for
> X86_PAE"
> was unintentionally removed by the recent
> "34437e67a672: x86/mm: Fix slow_virt_to_phys() to handle large PAT bit".
>
> And, the variable 'phys_addr' was defined as "unsigned long" by mistake
> -- it should
> be "phys_addr_t".
>
> As a result, Hyper-V network driver in 32-PAE Linux guest can't work
> again.
>
> Fixes: "commmit 34437e67a672: x86/mm: Fix slow_virt_to_phys() to handle
> large PAT bit"
> Signed-off-by: Dexuan Cui <decui@xxxxxxxxxxxxx>
> Cc: Toshi Kani <toshi.kani@xxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: K. Y. Srinivasan <kys@xxxxxxxxxxxxx>
> Cc: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
> Cc: gregkh@xxxxxxxxxxxxxxxxxxx
> Cc: linux-mm@xxxxxxxxx
> Cc: olaf@xxxxxxxxx
> Cc: apw@xxxxxxxxxxxxx
> Cc: jasowang@xxxxxxxxxx
> Cc: stable@xxxxxxxxxxxxxxx

Thanks for the fix and adding the comment to explain the trick!  The change
looks good to me.

Reviewed-by: Toshi Kani <toshi.kani@xxxxxxx>

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1549601

Title:
  [Hyper-V] x86,pageattr: prevent overflow in slow_virt_to_phys() for
  X86_PAE

Status in linux package in Ubuntu:
  Triaged

Bug description:
  https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=d1cd1210834649ce1ca6bafe5ac25d2f40331343

  x86, pageattr: Prevent overflow in slow_virt_to_phys() for X86_PAE
  pte_pfn() returns a PFN of long (32 bits in 32-PAE), so "long <<
  PAGE_SHIFT" will overflow for PFNs above 4GB.

  Due to this issue, some Linux 32-PAE distros, running as guests on Hyper-V,
  with 5GB memory assigned, can't load the netvsc driver successfully and
  hence the synthetic network device can't work (we can use the kernel parameter
  mem=3000M to work around the issue).

  Cast pte_pfn() to phys_addr_t before shifting.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1549601/+subscriptions