Skip to contentSkip to conversation

mremap: skip page table lookup for non-faulted anonymous VMAs

PATCH In review Nai Xia 2 participants 3 messages Nov 7, 2011

Subsystems (2)
lore
Share & tools
lore

Nai Xia

author
#0 lore raw

PATCH 1/1 mremap: skip page table lookup for non-faulted anonymous VMAs

If an anonymous vma has not yet been faulted, move_page_tables() in move_vma() is not necessary for it.

---

mm/mremap.c +5 -0
@@ -200,6 +200,10 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (!new_vma)
 		return -ENOMEM;
 
+	/* An anonymous vma has not been faulted, no pagetables lookup. */
+	if (!vma->vm_file && !vma->anon_vma)
+		goto page_tables_ok;
+
 	moved_len = move_page_tables(vma, old_addr, new_vma, new_addr, old_len);
 	if (moved_len < old_len) {
 		/*
@@ -213,6 +217,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 		old_addr = new_addr;
 		new_addr = -ENOMEM;
 	}
+page_tables_ok:
 
 	/* Conceal VM_ACCOUNT so old reservation is not undone */
 	if (vm_flags & VM_ACCOUNT) {

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#1 lore raw
Quoted 2 lines · Nai Xia

If an anonymous vma has not yet been faulted, move_page_tables() in move_vma() is not necessary for it.

I actually thought of adding that (in fact fork has it and it's more likely to be beneficial for fork than for mremap I suspect), but this adds a branch to the fast path for a case that shouldn't normally materialize. So I don't think it's worth adding it as I expect it to add overhead in average.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

author
#2 lore raw
Quoted 10 lines · Andrea Arcangeli
Quoted 2 lines · Nai Xia

If an anonymous vma has not yet been faulted, move_page_tables() in move_vma() is not necessary for it.

I actually thought of adding that (in fact fork has it and it's more likely to be beneficial for fork than for mremap I suspect), but this adds a branch to the fast path for a case that shouldn't normally materialize. So I don't think it's worth adding it as I expect it to add overhead in average.

Well, it seems like I forgot to embrace it with unlikely(), with instr prefetching this line seems a trivial even for the average cases.

But in case it really materializes, it may avoid tedious page table locking and pmd allocation and furthermore, a semantically confusing operation of move_page_tables() on a same VMA.

Anyway, it's not a big deal, indeed. You are the maintainers, it's left for you to decide.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>