Skip to contentSkip to conversation

powerpc/mm/hugetlb: Filter out hugepage size not supported by page table layout

PATCH In review Aneesh Kumar K.V 2 participants 5 messages Feb 19, 2017 – Feb 20, 2017 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc/mm/hugetlb: Filter out hugepage size not supported by page table layout

Without this if firmware reports 1MB page size support we will crash trying to use 1MB as hugetlb page size.

echo 300 > /sys/kernel/mm/hugepages/hugepages-1024kB/nr_hugepages

kernel BUG at ./arch/powerpc/include/asm/hugetlb.h:19!
.....
....
[c0000000e2c27b30] c00000000029dae8 .hugetlb_fault+0x638/0xda0 [c0000000e2c27c30] c00000000026fb64 .handle_mm_fault+0x844/0x1d70 [c0000000e2c27d70] c00000000004805c .do_page_fault+0x3dc/0x7c0 [c0000000e2c27e30] c00000000000ac98 handle_page_fault+0x10/0x30

With fix, we don't enable 1MB as hugepage size.

bash-4.2# cd /sys/kernel/mm/hugepages/
bash-4.2# ls
hugepages-16384kB hugepages-16777216kB

---
 arch/powerpc/mm/hugetlbpage.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
arch/powerpc/mm/hugetlbpage.c +18 -0
index 8c3389cbcd12..a4f33de4008e 100644
@@ -753,6 +753,24 @@ static int __init add_huge_page_size(unsigned long long size)
 	if ((mmu_psize = shift_to_mmu_psize(shift)) < 0)
 		return -EINVAL;
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	/*
+	 * We need to make sure that for different page sizes reported by
+	 * firmware we only add hugetlb support for page sizes that can be
+	 * supported by linux page table layout.
+	 * For now we have
+	 * Radix: 2M
+	 * Hash: 16M and 16G
+	 */
+	if (radix_enabled()) {
+		if (mmu_psize != MMU_PAGE_2M)
+			return -EINVAL;
+	} else {
+		if (mmu_psize != MMU_PAGE_16M && mmu_psize != MMU_PAGE_16G)
+			return -EINVAL;
+	}
+#endif
+
 	BUG_ON(mmu_psize_defs[mmu_psize].shift != shift);
 
 	/* Return if huge page size has already been setup */
Signature · 1 line
2.7.4
#1 lore raw
Quoted 8 lines · Aneesh Kumar K.V

+#ifdef CONFIG_PPC_BOOK3S_64
+       /*
+        * We need to make sure that for different page sizes reported by +        * firmware we only add hugetlb support for page sizes that can be +        * supported by linux page table layout.
+        * For now we have
+        * Radix: 2M
+        * Hash: 16M and 16G

Quoted 8 lines

+        */
+       if (radix_enabled()) {
+               if (mmu_psize != MMU_PAGE_2M)
+                       return -EINVAL;
+       } else {
+               if (mmu_psize != MMU_PAGE_16M && mmu_psize != MMU_PAGE_16G) +                       return -EINVAL;
+       }

Hash could support others... Same with radix and PUD level pages.

Why do we need that ? Won't FW provide separate properties for hash and radix page sizes anyway ?

Ben.

#2 lore raw
Quoted 17 lines · Benjamin Herrenschmidt
Quoted 6 lines · Aneesh Kumar K.V
arch/powerpc/mm/hugetlbpage.c +6 -0
+#ifdef CONFIG_PPC_BOOK3S_64
+       /*
+        * We need to make sure that for different page sizes reported by
+        * firmware we only add hugetlb support for page sizes that can be
+        * supported by linux page table layout.
+        * For now we have

arch/powerpc/mm/hugetlbpage.c static int __init add_huge_page_size(unsigned long long size) · line 771

Quoted 10 lines
arch/powerpc/mm/hugetlbpage.c +10 -0
+        * Radix: 2M
+        * Hash: 16M and 16G
+        */
+       if (radix_enabled()) {
+               if (mmu_psize != MMU_PAGE_2M)
+                       return -EINVAL;
+       } else {
+               if (mmu_psize != MMU_PAGE_16M && mmu_psize != MMU_PAGE_16G)
+                       return -EINVAL;
+       }
Quoted 1 line

Hash could support others...

On book3s 64 ? I had the above within #ifdef.

Quoted 1 line

Same with radix and PUD level pages.

Yes, but gigantic hugepage is not yet supported. Once we add that we will add MMU_PAGE_1G here.

Quoted 4 lines

Why do we need that ? Won't FW provide separate properties for hash and radix page sizes anyway ?

To avoid crashes like the one reported in the commit message due to buggy firmware ? Also It can serve as an easy way to understand what hugepage sizes are supported by different platforms. I am yet to figure out what the FSL_BOOK3E and PPC_8xx #ifdef above that hunk is all about. Having the supported hugepage size clearly verified against makes it easy ?

-aneesh

#3 lore raw
Quoted 2 lines · Aneesh Kumar K.V

To avoid crashes like the one reported in the commit message due to buggy firmware ?

I don't want Linux to make those assumptions. We should fix the FW.

Think of backward compat for example.

Quoted 9 lines

Also
It can serve as an easy way to understand what hugepage sizes are supported by different platforms.
I am yet to figure out what the FSL_BOOK3E and PPC_8xx #ifdef above that
hunk is all about. Having
the supported hugepage size clearly verified against makes it easy ?

-aneesh

#4 lore raw
Quoted 6 lines · Benjamin Herrenschmidt
Quoted 2 lines · Aneesh Kumar K.V

To avoid crashes like the one reported in the commit message due to=C2=A0 buggy firmware ?=20

I don't want Linux to make those assumptions. We should fix the FW.

I was not suggesting to not fix FW. The idea was two fold.

We cannot support different hugetlb page sizes. They need to be supported at linux page table level. So a generic check like is_power_of_2/4() may not be what we want. The second is to document clearly what are the different page sizes supported by a platform.

-aneesh