Re: [PATCH RFC] mm: account VMA before forced-COW via /proc/pid/mem

[Oleg Nesterov]

On 04/02, Konstantin Khlebnikov wrote:
>
> Currently kernel does not account read-only private mappings into memory commitment.
> But these mappings can be force-COW-ed in get_user_pages().

Heh. tail -n3 Documentation/vm/overcommit-accounting
may be you should update it then.

Can't really comment the patch, this is not my area. Still,

> +	down_write(&mm->mmap_sem);
> +	*pvma = vma = find_vma(mm, addr);
> +	if (vma && vma->vm_start <= addr) {
> +		ret = vma->vm_end - addr;
> +		if ((vma->vm_flags & (VM_ACCOUNT | VM_NORESERVE | VM_SHARED |
> +				VM_HUGETLB | VM_MAYWRITE)) == VM_MAYWRITE) {
> +			if (!security_vm_enough_memory_mm(mm, vma_pages(vma)))

Oooooh, the whole vma. Say, gdb installs the single breakpoint into
the huge .text mapping...

I am not sure, but probably you want to check at least VM_IO/PFNMAP
as well. We do not want to charge this memory and retry with FOLL_FORCE
before vm_ops->access(). Say, /dev/mem.

Hmm. OTOH, if I am right then mprotect_fixup() should be fixed??


We drop ->mmap_sem... Say, the task does mremap() in between and
len == 2 * PAGE_SIZE. Then, for example, copy_to_user_page() can
write to the same page twice. Perhaps not a problem in practice,
I dunno.

Oleg.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/