11:37 p.m.
On 04/07/2015 04:26 PM, Kirill A. Shutemov wrote:
On Tue, Apr 07, 2015 at 04:17:00PM +0300, Kirill A. Shutemov wrote:
> On Tue, Apr 07, 2015 at 03:57:32PM +0300, Boaz Harrosh wrote:
>> +/*
>> + * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
>> + * mapping
>> + */
>> +static int wp_pfn_shared(struct mm_struct *mm,
>> + struct vm_area_struct *vma, unsigned long address,
>> + pte_t *page_table, spinlock_t *ptl, pte_t orig_pte,
>> + pmd_t *pmd)
>> +{
>> + if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
>> + struct vm_fault vmf = {
>> + .page = NULL,
>> + .pgoff = linear_page_index(vma, address),
>> + .virtual_address = (void __user *)(address & PAGE_MASK),
>> + .flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE,
>> + };
>> + int ret;
>> +
>> + pte_unmap_unlock(page_table, ptl);
>> + ret = vma->vm_ops->pfn_mkwrite(vma, &vmf);
>> + if (ret & VM_FAULT_ERROR)
>> + return ret;
>> + page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
>> + /* Did pfn_mkwrite already fixed up the pte */
Oh. I guess you've missunderstood why we need pte_same() check below.
It's not about ->pfn_mkwrite() changing the pte (generatlly, it should
not). It's requited to address race with parallel page fault to the pte.
>> + if (!pte_same(*page_table, orig_pte)) {
>> + pte_unmap_unlock(page_table, ptl);
>> + return ret;
>
> This should be "return 0;", shouldn't it?
>
> VM_FAULT_NOPAGE would imply you've installed new pte, but you did not.
Changing this to "return 0" would be very scary for me. Because I'm running
with this code for 1/2 a year now. And it is stable. You see since the original
code it was always doing just that pte_unmap_unlock && return ret. (See the patch
based on 4.0)
I did not understand if you want that I keep it "return ret".
I gather that you would like the comment changed, about the changed pte.
Both here and at Documentation/.../locking.
I'll send a new patch just tell me if you want the reurn thing
Thank you
Boaz