Re: [PATCH v1 3/5] mm: ptdump: Provide page size to notepage()

From: Christophe Leroy
Date: Mon Apr 19 2021 - 12:41:22 EST




Le 19/04/2021 à 16:00, Steven Price a écrit :
On 19/04/2021 14:14, Christophe Leroy wrote:


Le 16/04/2021 à 12:51, Steven Price a écrit :
On 16/04/2021 11:38, Christophe Leroy wrote:


Le 16/04/2021 à 11:28, Steven Price a écrit :
On 15/04/2021 18:18, Christophe Leroy wrote:

To be honest I don't fully understand why powerpc requires the page_size - it appears to be using it purely to find "holes" in the calls to note_page(), but I haven't worked out why such holes would occur.

I was indeed introduced for KASAN. We have a first commit https://github.com/torvalds/linux/commit/cabe8138 which uses page size to detect whether it is a KASAN like stuff.

Then came https://github.com/torvalds/linux/commit/b00ff6d8c as a fix. I can't remember what the problem was exactly, something around the use of hugepages for kernel memory, came as part of the series https://patchwork.ozlabs.org/project/linuxppc-dev/cover/cover.1589866984.git.christophe.leroy@xxxxxxxxxx/




Ah, that's useful context. So it looks like powerpc took a different route to reducing the KASAN output to x86.

Given the generic ptdump code has handling for KASAN already it should be possible to drop that from the powerpc arch code, which I think means we don't actually need to provide page size to notepage(). Hopefully that means more code to delete ;)


Looking at how the generic ptdump code handles KASAN, I'm a bit sceptic.

IIUC, it is checking that kasan_early_shadow_pte is in the same page as the pgtable referred by the PMD entry. But what happens if that PMD entry is referring another pgtable which is inside the same page as kasan_early_shadow_pte ?

Shouldn't the test be

     if (pmd_page_vaddr(val) == lm_alias(kasan_early_shadow_pte))
         return note_kasan_page_table(walk, addr);

Now I come to look at this code again, I think you're right. On arm64 this doesn't cause a problem - page tables are page sized and page aligned, so there couldn't be any non-KASAN pgtables sharing the page. Obviously that's not necessarily true of other architectures.

Feel free to add a patch to your series ;)


Ok.

I'll leave that outside of the series, it is not a show stopper because early shadow page directories are all tagged __bss_page_aligned so we can't have two of them in the same page and it is really unlikely that we'll have any other statically defined page directory in the same pages either.

And for the special case of powerpc 8xx which is the only one for which we have both KASAN and HUGEPD at the time being, there are only two levels of page directories so no issue.

Christophe