Re: [PATCH] Revert "mm/usercopy: Drop extra is_vmalloc_or_module() check"
From: Kefeng Wang <hidden>
Date: 2021-12-25 12:00:26
Also in:
linuxppc-dev, lkml
On 2021/12/25 19:04, Nicholas Piggin wrote:
Excerpts from Kefeng Wang's message of December 25, 2021 12:05 pm:
...
quoted
quoted
Can you try that ? #define virt_addr_valid(kaddr) ((kaddr & PAGE_OFFSET) == PAGE_OFFSET && pfn_valid(virt_to_pfn(kaddr)))I got this commit, commit 4dd7554a6456d124c85e0a4ad156625b71390b5c Author: Nicholas Piggin [off-list ref] Date: Wed Jul 24 18:46:37 2019 +1000 powerpc/64: Add VIRTUAL_BUG_ON checks for __va and __pa addresses Ensure __va is given a physical address below PAGE_OFFSET, and __pa is given a virtual address above PAGE_OFFSET. It has check the PAGE_OFFSET in __pa, will test it and resend the patch(with above warning changes).What did you get with this commit? Is this what causes the crash?
I mean that your patch does the check to make sure the virt addr should above PAGE_OFFSET, and we can add the check in the virt_addr_valid too.
quoted hunk ↗ jump to hunk
riscv for example with flatmem also relies on pfn_valid to do the right thing, so as far as I can see the check should exclude vmalloc addresses and it's just a matter of virt_addr_valid not to give virt_to_pfn an address < PAGE_OFFSET. If we take riscv's implementationdiff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h index 254687258f42..7713188516a6 100644 --- a/arch/powerpc/include/asm/page.h +++ b/arch/powerpc/include/asm/page.h@@ -132,7 +132,10 @@ static inline bool pfn_valid(unsigned long pfn) #define virt_to_page(kaddr) pfn_to_page(virt_to_pfn(kaddr)) #define pfn_to_kaddr(pfn) __va((pfn) << PAGE_SHIFT) -#define virt_addr_valid(kaddr) pfn_valid(virt_to_pfn(kaddr)) +#define virt_addr_valid(vaddr) ({ \ + unsigned long _addr = (unsigned long)vaddr; \ + (unsigned long)(_addr) >= PAGE_OFFSET && pfn_valid(virt_to_pfn(_addr)); \ +})
Yes, I send a new v2 with this change, thanks
/* * On Book-E parts we need __va to parse the device tree and we can't .