Skip to contentSkip to conversation

mm: Allow shmem mappings with MREMAP_DONTUNMAP

PATCH In review Brian Geffon 8 participants 34 messages Mar 3, 2021 – May 10, 2022

Subsystems (8)
Memory management x86 VFS Kselftest +4 more subsystems
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm: Allow shmem mappings with MREMAP_DONTUNMAP

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include shmem mappings. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases. With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/mremap.c +1 -2
index ec8f840399ed..6934d199da54 100644
@@ -653,8 +653,7 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if (flags & MREMAP_DONTUNMAP && !(vma_is_anonymous(vma) || vma_is_shmem(vma)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))
Signature · 1 line
2.31.0.rc0.254.gbdcc3b1a9d-goog
#1 lore raw

I apologize, this patch didn't include my signed off by, here it is:

Quoted 49 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include shmem mappings. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases. With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/mremap.c +1 -3
index ec8f840399ed..6934d199da54 100644
@@ -653,8 +653,7 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
                return ERR_PTR(-EINVAL);
        }
 
-       if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-                       vma->vm_flags & VM_SHARED))
+       if (flags & MREMAP_DONTUNMAP && !(vma_is_anonymous(vma) || vma_is_shmem(vma)))
                return ERR_PTR(-EINVAL);
 
        if (is_vm_hugetlb_page(vma))
--

2.31.0.rc0.254.gbdcc3b1a9d-goog

Tested-by
#2 lore raw
Quoted 35 lines · Brian Geffon

I apologize, this patch didn't include my signed off by, here it is:

Quoted 28 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include shmem mappings. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases. With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

Thanks for the patch, Brian. I've tested mremap(MREMAP_DONTUNMAP) on a memfd memory range.

Quoted 21 lines
Quoted 21 lines
---
 mm/mremap.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/mremap.c +1 -3
index ec8f840399ed..6934d199da54 100644
@@ -653,8 +653,7 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
                return ERR_PTR(-EINVAL);
        }
 
-       if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-                       vma->vm_flags & VM_SHARED))
+       if (flags & MREMAP_DONTUNMAP && !(vma_is_anonymous(vma) || vma_is_shmem(vma)))
                return ERR_PTR(-EINVAL);
 
        if (is_vm_hugetlb_page(vma))
--

2.31.0.rc0.254.gbdcc3b1a9d-goog

#3 lore raw
In reply to Brian Geffon
Quoted 37 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include shmem mappings. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases. With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/mremap.c +0 -0
index ec8f840399ed..6934d199da54 100644

mm/mremap.c static struct vm_area_struct *vma_to_resize(unsigned long addr, · line 659

Quoted 9 lines
mm/mremap.c +1 -2
@@ -653,8 +653,7 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if (flags & MREMAP_DONTUNMAP && !(vma_is_anonymous(vma) || vma_is_shmem(vma)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))

Yet something to improve...

Thanks for extending MREMAP_DONTUNMAP to shmem, but I think this patch goes in the wrong direction, complicating when it should be generalizing:
the mremap syscall is about rearranging the user's virtual address space, and is not specific to the underlying anonymous or shmem or file object (though so far you have only been interested in anonymous, and now shmem).

A better patch would say:

- if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
- vma->vm_flags & VM_SHARED))
+ if ((flags & MREMAP_DONTUNMAP) &&
+ (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
return ERR_PTR(-EINVAL);

VM_DONTEXPAND is what has long been used on special mappings, to prevent surprises from mremap changing the size of the mapping: MREMAP_DONTUNMAP introduced a different way of expanding the mapping, so VM_DONTEXPAND still seems a reasonable name (I've thrown in VM_PFNMAP there because it's in the VM_DONTEXPAND test lower down: for safety I guess, and best if both behave the same - though one says -EINVAL and the other -EFAULT).

With that VM_DONTEXPAND check in, Dmitry's commit cd544fd1dc92 ("mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio") can still be reverted (as you agreed on 28th December), even though vma_is_anonymous() will no longer protect it.

Was there an mremap(2) man page update for MREMAP_DONTUNMAP? Whether or not there was before, it ought to get one now.

Thanks,
Hugh

#4 lore raw

Hi Hugh,
Thanks for this suggestion, responses in line.

Quoted 8 lines

A better patch would say:

mm/mremap.c +2 -2
-       if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-                       vma->vm_flags & VM_SHARED))
+       if ((flags & MREMAP_DONTUNMAP) &&
+           (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
                return ERR_PTR(-EINVAL);
Quoted 6 lines

VM_DONTEXPAND is what has long been used on special mappings, to prevent surprises from mremap changing the size of the mapping: MREMAP_DONTUNMAP introduced a different way of expanding the mapping, so VM_DONTEXPAND still seems a reasonable name (I've thrown in VM_PFNMAP there because it's in the VM_DONTEXPAND test lower down: for safety I guess, and best if both behave the same - though one says -EINVAL and the other -EFAULT).

I like this idea and am happy to mail a new patch. I think it may make sense to bring the lower block up here so that it becomes more clear that it's not duplicate code and that the MREMAP_DONTUNMAP case returns -EINVAL and other cases return -EFAULT. I wonder if the -EFAULT error code would have made more sense from the start for both cases, do you have any thoughts on changing the error code at this point?

Quoted 4 lines

With that VM_DONTEXPAND check in, Dmitry's commit cd544fd1dc92 ("mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio") can still be reverted (as you agreed on 28th December), even though vma_is_anonymous() will no longer protect it.

I agree and if Dmitry does not have time I would be happy to mail a revert to cd544fd1dc92 as we discussed in [1]. Dmitry, would you like me to do that?

Quoted 2 lines

Was there an mremap(2) man page update for MREMAP_DONTUNMAP? Whether or not there was before, it ought to get one now.

Yes, the mremap(2) man page was updated when this flag was added and it will require a further update to reflect this expanded mapping support.

Thanks
Brian

1. https://lkml.org/lkml/2020/12/28/2340

#5 lore raw

Hi Brian, Hugh,

Quoted 26 lines · Brian Geffon

Hi Hugh,
Thanks for this suggestion, responses in line.

Quoted 8 lines

A better patch would say:

mm/mremap.c +2 -2
-       if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-                       vma->vm_flags & VM_SHARED))
+       if ((flags & MREMAP_DONTUNMAP) &&
+           (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
                return ERR_PTR(-EINVAL);
Quoted 6 lines

VM_DONTEXPAND is what has long been used on special mappings, to prevent surprises from mremap changing the size of the mapping: MREMAP_DONTUNMAP introduced a different way of expanding the mapping, so VM_DONTEXPAND still seems a reasonable name (I've thrown in VM_PFNMAP there because it's in the VM_DONTEXPAND test lower down: for safety I guess, and best if both behave the same - though one says -EINVAL and the other -EFAULT).

I like this idea and am happy to mail a new patch. I think it may make sense to bring the lower block up here so that it becomes more clear that it's not duplicate code and that the MREMAP_DONTUNMAP case returns -EINVAL and other cases return -EFAULT. I wonder if the -EFAULT error code would have made more sense from the start for both cases, do you have any thoughts on changing the error code at this point?

Quoted 8 lines
Quoted 4 lines

With that VM_DONTEXPAND check in, Dmitry's commit cd544fd1dc92 ("mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio") can still be reverted (as you agreed on 28th December), even though vma_is_anonymous() will no longer protect it.

I agree and if Dmitry does not have time I would be happy to mail a revert to cd544fd1dc92 as we discussed in [1]. Dmitry, would you like me to do that?

Ack. I was planning to send a patches set that includes the revert, but that's stalled a bit. As the patch just adds excessive checks, but doesn't introduce an issue, I haven't sent it separately. Feel free to revert it :-)

Thanks,
Dmitry

#6 lore raw
In reply to Hugh Dickins
Quoted 72 lines · Hugh Dickins
Quoted 37 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include shmem mappings. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases. With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
mm/mremap.c +0 -0
index ec8f840399ed..6934d199da54 100644

mm/mremap.c static struct vm_area_struct *vma_to_resize(unsigned long addr, · line 659

Quoted 9 lines
mm/mremap.c +1 -2
@@ -653,8 +653,7 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if (flags & MREMAP_DONTUNMAP && !(vma_is_anonymous(vma) || vma_is_shmem(vma)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))

Yet something to improve...

Thanks for extending MREMAP_DONTUNMAP to shmem, but I think this patch goes in the wrong direction, complicating when it should be generalizing:
the mremap syscall is about rearranging the user's virtual address space, and is not specific to the underlying anonymous or shmem or file object (though so far you have only been interested in anonymous, and now shmem).

A better patch would say:

mm/mremap.c +2 -2
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if ((flags & MREMAP_DONTUNMAP) &&
+	    (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
 		return ERR_PTR(-EINVAL);

VM_DONTEXPAND is what has long been used on special mappings, to prevent surprises from mremap changing the size of the mapping: MREMAP_DONTUNMAP introduced a different way of expanding the mapping, so VM_DONTEXPAND still seems a reasonable name (I've thrown in VM_PFNMAP there because it's in the VM_DONTEXPAND test lower down: for safety I guess, and best if both behave the same - though one says -EINVAL and the other -EFAULT).

Quoted 7 lines

With that VM_DONTEXPAND check in, Dmitry's commit cd544fd1dc92 ("mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio") can still be reverted (as you agreed on 28th December), even though vma_is_anonymous() will no longer protect it.

Was there an mremap(2) man page update for MREMAP_DONTUNMAP? Whether or not there was before, it ought to get one now.

I'm curious whether it's okay to expand MREMAP_DONTUNMAP to PFNMAP too.. E.g. vfio maps device MMIO regions with both VM_DONTEXPAND|VM_PFNMAP, to me it makes sense to allow the userspace to get such MMIO region remapped/duplicated somewhere else as long as the size won't change. With the strict check as above we kill all those possibilities.

Though in that case we'll still need commits like cd544fd1dc92 to protect any customized ->mremap() when they're not supported.

Thanks,

Signature · 1 line
Peter Xu
#7 lore raw
Quoted 9 lines · Peter Xu

I'm curious whether it's okay to expand MREMAP_DONTUNMAP to PFNMAP too.. E.g. vfio maps device MMIO regions with both VM_DONTEXPAND|VM_PFNMAP, to me it makes sense to allow the userspace to get such MMIO region remapped/duplicated somewhere else as long as the size won't change. With the strict check as above we kill all those possibilities.

Though in that case we'll still need commits like cd544fd1dc92 to protect any customized ->mremap() when they're not supported.

It would take me many hours to arrive at a conclusion on that:
I'm going to spend the time differently, and let whoever ends up wanting MREMAP_DONTUNMAP on a VM_PFNMAP area research the safety of that for existing users.

I did look to see what added VM_PFNMAP to the original VM_DONTEXPAND:

v2.6.15
commit 4d7672b46244abffea1953e55688c0ea143dd617

Author: Linus Torvalds <[email protected]>
Date:   Fri Dec 16 10:21:23 2005 -0800

Make sure we copy pages inserted with "vm_insert_page()" on fork

The logic that decides that a fork() might be able to avoid copying a VM area when it can be re-created by page faults didn't know about the new vm_insert_page() case.

Also make some things a bit more anal wrt VM_PFNMAP.

Pointed out by Hugh Dickins <[email protected]>

So apparently I do bear some anal responsibility. My concern seems to have been that in those days an unexpected page fault in a special driver area would end up allocating an anonymous page, which would never get freed later. Nowadays it looks like there's a SIGBUS for the equivalent situation.

So probably VM_DONTEXPAND is less important than it was, and the additional VM_PFNMAP safety net no longer necessary, and you could strip it out of the old size check and Brian's new dontunmap check.

But I give no guarantee: I don't know VM_PFNMAP users at all well.

Hugh

#8 lore raw

PATCH v2 1/2 mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

In reply to Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change which takes advantage of the existing check in vma_to_resize for non-VM_DONTEXPAND and non-VM_PFNMAP mappings will cause MREMAP_DONTUNMAP to return -EFAULT if such mappings are remapped. This behavior is consistent with existing behavior when using mremap with such mappings.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ----
 1 file changed, 4 deletions(-)
mm/mremap.c +0 -4
index ec8f840399ed..2c57dc4bc8b6 100644
@@ -653,10 +653,6 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
-		return ERR_PTR(-EINVAL);
-
 	if (is_vm_hugetlb_page(vma))
 		return ERR_PTR(-EINVAL);
 
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#9 lore raw

PATCH v2 2/2 Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 77e64e3eac80..8c3729eb3e38 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index 2c57dc4bc8b6..b1f7bc43ece9 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#10 lore raw

Re: [PATCH v2 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

In reply to Brian Geffon

Hi, Brian,

Quoted 49 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change which takes advantage of the existing check in vma_to_resize for non-VM_DONTEXPAND and non-VM_PFNMAP mappings will cause MREMAP_DONTUNMAP to return -EFAULT if such mappings are remapped. This behavior is consistent with existing behavior when using mremap with such mappings.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ----
 1 file changed, 4 deletions(-)
mm/mremap.c +0 -0
index ec8f840399ed..2c57dc4bc8b6 100644

mm/mremap.c static struct vm_area_struct *vma_to_resize(unsigned long addr, · line 657

Quoted 9 lines
mm/mremap.c +0 -4
@@ -653,10 +653,6 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
-		return ERR_PTR(-EINVAL);
-
 	if (is_vm_hugetlb_page(vma))
 		return ERR_PTR(-EINVAL);

The code change seems to be not aligned with what the commit message said. Did you perhaps forget to add the checks against VM_DONTEXPAND | VM_PFNMAP? I'm guessing that (instead of commit message to be touched up) because you still attached the revert patch, then that check seems to be needed. Thanks,

Signature · 1 line
Peter Xu
#11 lore raw

Re: [PATCH v2 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

Hi Peter,
Thank you as always for taking a look. This change relies on the existing check in vma_to_resize on line 686:
https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L686
which returns -EFAULT when the vma is VM_DONTEXPAND or VM_PFNMAP.

Thanks
Brian

Quoted 71 lines · Peter Xu

Hi, Brian,

Quoted 49 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change which takes advantage of the existing check in vma_to_resize for non-VM_DONTEXPAND and non-VM_PFNMAP mappings will cause MREMAP_DONTUNMAP to return -EFAULT if such mappings are remapped. This behavior is consistent with existing behavior when using mremap with such mappings.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ----
 1 file changed, 4 deletions(-)
mm/mremap.c +0 -0
index ec8f840399ed..2c57dc4bc8b6 100644

mm/mremap.c static struct vm_area_struct *vma_to_resize(unsigned long addr, · line 657

Quoted 9 lines
mm/mremap.c +0 -4
@@ -653,10 +653,6 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
              return ERR_PTR(-EINVAL);
      }
 
-     if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-                     vma->vm_flags & VM_SHARED))
-             return ERR_PTR(-EINVAL);
-
      if (is_vm_hugetlb_page(vma))
              return ERR_PTR(-EINVAL);

The code change seems to be not aligned with what the commit message said. Did you perhaps forget to add the checks against VM_DONTEXPAND | VM_PFNMAP? I'm guessing that (instead of commit message to be touched up) because you still attached the revert patch, then that check seems to be needed. Thanks,

--
Peter Xu

#12 lore raw

Re: [PATCH v2 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

Quoted 1 line · Brian Geffon

Hi Peter,

Hi, Brian,

Quoted 4 lines

Thank you as always for taking a look. This change relies on the existing check in vma_to_resize on line 686:
https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L686
which returns -EFAULT when the vma is VM_DONTEXPAND or VM_PFNMAP.

Do you mean line 676?

https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L676

I'm not sure whether it'll work for MREMAP_DONTUNMAP, since IIUC MREMAP_DONTUNMAP only works for the remap case with no size change, however in that case in vma_to_resize() we'll bail out even earlier than line 676 when checking against the size:

https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L667

So IIUC we'll still need the change as Hugh suggested previously.

Thanks,

Signature · 1 line
Peter Xu
#13 lore raw

Re: [PATCH v2 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

You're 100% correct, I'll mail a new patch in a few

Brian

Quoted 29 lines · Peter Xu
Quoted 1 line · Brian Geffon

Hi Peter,

Hi, Brian,

Quoted 4 lines

Thank you as always for taking a look. This change relies on the existing check in vma_to_resize on line 686:
https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L686
which returns -EFAULT when the vma is VM_DONTEXPAND or VM_PFNMAP.

Do you mean line 676?

https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L676

I'm not sure whether it'll work for MREMAP_DONTUNMAP, since IIUC MREMAP_DONTUNMAP only works for the remap case with no size change, however in that case in vma_to_resize() we'll bail out even earlier than line 676 when checking against the size:

https://elixir.bootlin.com/linux/v5.12-rc3/source/mm/mremap.c#L667

So IIUC we'll still need the change as Hugh suggested previously.

Thanks,

--
Peter Xu

#14 lore raw

PATCH v3 1/2 mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

In reply to Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
mm/mremap.c +2 -2
index ec8f840399ed..db5b8b28c2dd 100644
@@ -653,8 +653,8 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if ((flags & MREMAP_DONTUNMAP) &&
+			(vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#15 lore raw

PATCH v3 2/2 Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 77e64e3eac80..8c3729eb3e38 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
Acked-by on v3 2/2
#16 lore raw

Re: [PATCH v3 2/2] Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

Quoted 8 lines · Brian Geffon

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

One small fixup needed: -EFAULT was what the incorrect v2 gave, but v3 issues -EINVAL like before, and I'm content with that difference.

Quoted 4 lines

1. https://lkml.org/lkml/2020/12/28/2340

Thanks Brian, I'm happy with this result.

Quoted 91 lines
---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 77e64e3eac80..8c3729eb3e38 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
maintainer
#17 lore raw

Re: [PATCH v3 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

In reply to Brian Geffon
Quoted 3 lines · On Wed, 17 Mar 2021 14:41:46 -0700 Brian Geffon <[email protected]> wrote:

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP.

Please update changelog to explain why these two were omitted?

Quoted 23 lines

The primary use case is to support
MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

Quoted 7 lines

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

Is a manpage update planned? It's appropriate to add this to the series so folks can check it over.

Can we please get the appropriate updates into tools/testing/selftests/vm/mremap_test.c for this?

Quoted 1 line

v3 is getting up there. Has there been much review activity?

Acked-by on v3 1/2
#18 lore raw

Re: [PATCH v3 1/2] mm: Allow non-VM_DONTEXPAND and VM_PFNMAP mappings with MREMAP_DONTUNMAP

In reply to Brian Geffon

If Andrew is happy with such a long patch name, okay; but personally I'd prefer brevity to all that detail:

mm: Extend MREMAP_DONTUNMAP to non-anonymous mappings

Quoted 33 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd.

This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

Quoted 4 lines

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

Thanks Brian, just what I wanted :)

You wondered in another mail about this returning -EINVAL whereas the VM_DONTEXPAND size error returns -EFAULT: I've pondered, and I've read the manpage, and I'm sure it would be wrong to change the old -EFAULT to -EINVAL now; and I don't see good reason to change your -EINVAL to -EFAULT either. Let them differ, that's okay (and it's only in special corner cases that either of these fail anyway).

Quoted 21 lines
---
 mm/mremap.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
mm/mremap.c +2 -2
index ec8f840399ed..db5b8b28c2dd 100644
@@ -653,8 +653,8 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if ((flags & MREMAP_DONTUNMAP) &&
+			(vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#19 lore raw

PATCH v4 1/3 mm: Extend MREMAP_DONTUNMAP to non-anonymous mappings

In reply to Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This restriction was placed initially for simplicity and not because there exists a technical reason to do so.

This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd. This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
mm/mremap.c +2 -2
index ec8f840399ed..db5b8b28c2dd 100644
@@ -653,8 +653,8 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if ((flags & MREMAP_DONTUNMAP) &&
+			(vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#20 lore raw

PATCH v4 2/3 Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 64a71bf20536..ecdc6e8dc5af 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#21 lore raw

Re: [PATCH v4 2/3] Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

Quoted 8 lines · Brian Geffon

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

No, those two lines still describe an earlier version, they should say:
"mappings, and these special mappings are now protected by a check of
!VM_DONTEXPAND and !VM_PFNMAP which will result in a -EINVAL."

Quoted 98 lines

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 64a71bf20536..ecdc6e8dc5af 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#22 lore raw

Re: [PATCH v4 2/3] Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

I apologize for that oversight, a new version with the updated commit message has been mailed.

Brian

Quoted 120 lines · Hugh Dickins
Quoted 8 lines · Brian Geffon

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EFAULT.

No, those two lines still describe an earlier version, they should say:
"mappings, and these special mappings are now protected by a check of
!VM_DONTEXPAND and !VM_PFNMAP which will result in a -EINVAL."

Quoted 13 lines

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)

Review

Quoted 1 line
(quoted diff) +0 -0

b/arch/x86/kernel/cpu/resctrl/pseudo_lock.c

arch/x86/kernel/cpu/resctrl/pseudo_lock.c static int pseudo_lock_dev_release(struct inode

Quoted 4 lines

index e916646adc69..0daf2f1cf7a8 100644

arch/x86/kernel/cpu/resctrl/pseudo_lock.c +0 -0
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode

*inode, struct file *filp)

arch/x86/kernel/cpu/resctrl/pseudo_lock.c static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)

Quoted 4 lines
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +0 -1
      return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned

long flags)

fs/aio.c static void aio_free_ring(struct kioctx *ctx) · line 326

Quoted 13 lines
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -0
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
      /* Not supported */
      return -EINVAL;
fs/aio.c +0 -1
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
      }
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long

flags)

mm/mmap.c static const char *special_mapping_name(struct

Quoted 31 lines
fs/aio.c +1 -3
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
      struct file *file = vma->vm_file;
      struct mm_struct *mm = vma->vm_mm;
      struct kioctx_table *table;
      int i, res = -EINVAL;
 
-     if (flags & MREMAP_DONTUNMAP)
-             return -EINVAL;
-
      spin_lock(&mm->ioctx_lock);
      rcu_read_lock();
      table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 64a71bf20536..ecdc6e8dc5af 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
      void (*close)(struct vm_area_struct * area);
      /* Called any time before splitting to check if it's allowed */
      int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-     int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+     int (*mremap)(struct vm_area_struct *area);
      /*
       * Called by mprotect() to make driver-specific permission
       * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +0 -0
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct

vm_area_struct *vma)

mm/mremap.c static unsigned long move_vma(struct vm_area_struct

Quoted 20 lines
mm/mmap.c +1 -5
      return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-                               unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
      struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-     if (flags & MREMAP_DONTUNMAP)
-             return -EINVAL;
-
      if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
              return -EFAULT;
 
mm/mremap.c +0 -0
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct

*vma,

Quoted 12 lines
mm/mremap.c +1 -2
      if (moved_len < old_len) {
              err = -ENOMEM;
      } else if (vma->vm_ops && vma->vm_ops->mremap) {
-             err = vma->vm_ops->mremap(new_vma, flags);
+             err = vma->vm_ops->mremap(new_vma);
      }
 
      if (unlikely(err)) {
--

2.31.0.rc2.261.g7f71774620-goog

#23 lore raw

PATCH v4 3/3 selftests: Add a MREMAP_DONTUNMAP selftest for shmem

In reply to Brian Geffon

This test extends the current mremap tests to validate that the MREMAP_DONTUNMAP operation can be performed on shmem mappings.

---
 tools/testing/selftests/vm/mremap_dontunmap.c | 52 +++++++++++++++++++
 1 file changed, 52 insertions(+)
tools/testing/selftests/vm/mremap_dontunmap.c +52 -0
index 3a7b5ef0b0c6..f01dc4a85b0b 100644
@@ -127,6 +127,57 @@ static void mremap_dontunmap_simple()
 	       "unable to unmap source mapping");
 }
 
+// This test validates that MREMAP_DONTUNMAP on a shared mapping works as expected.
+static void mremap_dontunmap_simple_shmem()
+{
+	unsigned long num_pages = 5;
+
+	int mem_fd = memfd_create("memfd", MFD_CLOEXEC);
+	BUG_ON(mem_fd < 0, "memfd_create");
+
+	BUG_ON(ftruncate(mem_fd, num_pages * page_size) < 0,
+			"ftruncate");
+
+	void *source_mapping =
+	    mmap(NULL, num_pages * page_size, PROT_READ | PROT_WRITE,
+		 MAP_FILE | MAP_SHARED, mem_fd, 0);
+	BUG_ON(source_mapping == MAP_FAILED, "mmap");
+
+	BUG_ON(close(mem_fd) < 0, "close");
+
+	memset(source_mapping, 'a', num_pages * page_size);
+
+	// Try to just move the whole mapping anywhere (not fixed).
+	void *dest_mapping =
+	    mremap(source_mapping, num_pages * page_size, num_pages * page_size,
+		   MREMAP_DONTUNMAP | MREMAP_MAYMOVE, NULL);
+	if (dest_mapping == MAP_FAILED && errno == EINVAL) {
+		// Old kernel which doesn't support MREMAP_DONTUNMAP on shmem.
+		BUG_ON(munmap(source_mapping, num_pages * page_size) == -1,
+			"unable to unmap source mapping");
+		return;
+	}
+
+	BUG_ON(dest_mapping == MAP_FAILED, "mremap");
+
+	// Validate that the pages have been moved, we know they were moved if
+	// the dest_mapping contains a's.
+	BUG_ON(check_region_contains_byte
+	       (dest_mapping, num_pages * page_size, 'a') != 0,
+	       "pages did not migrate");
+
+	// Because the region is backed by shmem, we will actually see the same
+	// memory at the source location still.
+	BUG_ON(check_region_contains_byte
+	       (source_mapping, num_pages * page_size, 'a') != 0,
+	       "source should have no ptes");
+
+	BUG_ON(munmap(dest_mapping, num_pages * page_size) == -1,
+	       "unable to unmap destination mapping");
+	BUG_ON(munmap(source_mapping, num_pages * page_size) == -1,
+	       "unable to unmap source mapping");
+}
+
 // This test validates MREMAP_DONTUNMAP will move page tables to a specific
 // destination using MREMAP_FIXED, also while validating that the source
 // remains intact.
@@ -300,6 +351,7 @@ int main(void)
 	BUG_ON(page_buffer == MAP_FAILED, "unable to mmap a page.");
 
 	mremap_dontunmap_simple();
+	mremap_dontunmap_simple_shmem();
 	mremap_dontunmap_simple_fixed();
 	mremap_dontunmap_partial_mapping();
 	mremap_dontunmap_partial_mapping_overwrite();
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#24 lore raw

PATCH 1/1 mremap.2: MREMAP_DONTUNMAP to reflect to supported mappings

In reply to Brian Geffon

mremap(2) now supports MREMAP_DONTUNMAP with mapping types other than private anonymous.

---
 man2/mremap.2 | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)
man2/mremap.2 +2 -11
index 3ed0c0c0a..72acbc111 100644
@@ -118,16 +118,6 @@ This flag, which must be used in conjunction with
 remaps a mapping to a new address but does not unmap the mapping at
 .IR old_address .
 .IP
-The
-.B MREMAP_DONTUNMAP
-flag can be used only with private anonymous mappings
-(see the description of
-.BR MAP_PRIVATE
-and
-.BR MAP_ANONYMOUS
-in
-.BR mmap (2)).
-.IP
 After completion,
 any access to the range specified by
 .IR old_address
@@ -227,7 +217,8 @@ was specified, but one or more pages in the range specified by
 .IR old_address
 and
 .IR old_size
-were not private anonymous;
+were part of a special mapping or the mapping is one that
+does not support merging or expanding;
 .IP *
 .B MREMAP_DONTUNMAP
 was specified and
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#25 lore raw

PATCH v5 1/3 mm: Extend MREMAP_DONTUNMAP to non-anonymous mappings

In reply to Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This restriction was placed initially for simplicity and not because there exists a technical reason to do so.

This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd. This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

---
 mm/mremap.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
mm/mremap.c +2 -2
index ec8f840399ed..db5b8b28c2dd 100644
@@ -653,8 +653,8 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
 		return ERR_PTR(-EINVAL);
 	}
 
-	if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
-			vma->vm_flags & VM_SHARED))
+	if ((flags & MREMAP_DONTUNMAP) &&
+			(vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)))
 		return ERR_PTR(-EINVAL);
 
 	if (is_vm_hugetlb_page(vma))
Signature · 1 line
2.31.0.291.g576ba9dcdaf-goog
#26 lore raw

PATCH v5 2/3 Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EINVAL.

1. https://lkml.org/lkml/2020/12/28/2340

---
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c | 2 +-
 fs/aio.c                                  | 5 +----
 include/linux/mm.h                        | 2 +-
 mm/mmap.c                                 | 6 +-----
 mm/mremap.c                               | 2 +-
 5 files changed, 5 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +1 -1
index e916646adc69..0daf2f1cf7a8 100644
@@ -1458,7 +1458,7 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area, unsigned long flags)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +1 -4
index 1f32da13d39e..76ce0cc3ee4e 100644
@@ -323,16 +323,13 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma, unsigned long flags)
+static int aio_ring_mremap(struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
 	int i, res = -EINVAL;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
 	table = rcu_dereference(mm->ioctx_table);
include/linux/mm.h +1 -1
index 64a71bf20536..ecdc6e8dc5af 100644
@@ -570,7 +570,7 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area, unsigned long flags);
+	int (*mremap)(struct vm_area_struct *area);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
mm/mmap.c +1 -5
index 3f287599a7a3..9d7651e4e1fe 100644
@@ -3403,14 +3403,10 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma,
-				  unsigned long flags)
+static int special_mapping_mremap(struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
-	if (flags & MREMAP_DONTUNMAP)
-		return -EINVAL;
-
 	if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))
 		return -EFAULT;
 
mm/mremap.c +1 -1
index db5b8b28c2dd..d22629ff8f3c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma, flags);
+		err = vma->vm_ops->mremap(new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.0.291.g576ba9dcdaf-goog
Reviewed-by on v5 2/3
#27 lore raw

Re: [PATCH v5 2/3] Revert "mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio"

Quoted 9 lines · Brian Geffon

This reverts commit cd544fd1dc9293c6702fab6effa63dac1cc67e99.

As discussed in [1] this commit was a no-op because the mapping type was checked in vma_to_resize before move_vma is ever called. This meant that vm_ops->mremap() would never be called on such mappings. Furthermore, we've since expanded support of MREMAP_DONTUNMAP to non-anonymous mappings, and these special mappings are still protected by the existing check of !VM_DONTEXPAND and !VM_PFNMAP which will result in a -EINVAL.

Quoted 4 lines

1. https://lkml.org/lkml/2020/12/28/2340

Thanks,
Dmitry

#28 lore raw

PATCH v5 3/3 selftests: Add a MREMAP_DONTUNMAP selftest for shmem

In reply to Brian Geffon

This test extends the current mremap tests to validate that the MREMAP_DONTUNMAP operation can be performed on shmem mappings.

---
 tools/testing/selftests/vm/mremap_dontunmap.c | 52 +++++++++++++++++++
 1 file changed, 52 insertions(+)
tools/testing/selftests/vm/mremap_dontunmap.c +52 -0
index 3a7b5ef0b0c6..f01dc4a85b0b 100644
@@ -127,6 +127,57 @@ static void mremap_dontunmap_simple()
 	       "unable to unmap source mapping");
 }
 
+// This test validates that MREMAP_DONTUNMAP on a shared mapping works as expected.
+static void mremap_dontunmap_simple_shmem()
+{
+	unsigned long num_pages = 5;
+
+	int mem_fd = memfd_create("memfd", MFD_CLOEXEC);
+	BUG_ON(mem_fd < 0, "memfd_create");
+
+	BUG_ON(ftruncate(mem_fd, num_pages * page_size) < 0,
+			"ftruncate");
+
+	void *source_mapping =
+	    mmap(NULL, num_pages * page_size, PROT_READ | PROT_WRITE,
+		 MAP_FILE | MAP_SHARED, mem_fd, 0);
+	BUG_ON(source_mapping == MAP_FAILED, "mmap");
+
+	BUG_ON(close(mem_fd) < 0, "close");
+
+	memset(source_mapping, 'a', num_pages * page_size);
+
+	// Try to just move the whole mapping anywhere (not fixed).
+	void *dest_mapping =
+	    mremap(source_mapping, num_pages * page_size, num_pages * page_size,
+		   MREMAP_DONTUNMAP | MREMAP_MAYMOVE, NULL);
+	if (dest_mapping == MAP_FAILED && errno == EINVAL) {
+		// Old kernel which doesn't support MREMAP_DONTUNMAP on shmem.
+		BUG_ON(munmap(source_mapping, num_pages * page_size) == -1,
+			"unable to unmap source mapping");
+		return;
+	}
+
+	BUG_ON(dest_mapping == MAP_FAILED, "mremap");
+
+	// Validate that the pages have been moved, we know they were moved if
+	// the dest_mapping contains a's.
+	BUG_ON(check_region_contains_byte
+	       (dest_mapping, num_pages * page_size, 'a') != 0,
+	       "pages did not migrate");
+
+	// Because the region is backed by shmem, we will actually see the same
+	// memory at the source location still.
+	BUG_ON(check_region_contains_byte
+	       (source_mapping, num_pages * page_size, 'a') != 0,
+	       "source should have no ptes");
+
+	BUG_ON(munmap(dest_mapping, num_pages * page_size) == -1,
+	       "unable to unmap destination mapping");
+	BUG_ON(munmap(source_mapping, num_pages * page_size) == -1,
+	       "unable to unmap source mapping");
+}
+
 // This test validates MREMAP_DONTUNMAP will move page tables to a specific
 // destination using MREMAP_FIXED, also while validating that the source
 // remains intact.
@@ -300,6 +351,7 @@ int main(void)
 	BUG_ON(page_buffer == MAP_FAILED, "unable to mmap a page.");
 
 	mremap_dontunmap_simple();
+	mremap_dontunmap_simple_shmem();
 	mremap_dontunmap_simple_fixed();
 	mremap_dontunmap_partial_mapping();
 	mremap_dontunmap_partial_mapping_overwrite();
Signature · 1 line
2.31.0.291.g576ba9dcdaf-goog
#29 lore raw

PATCH 1/1 mremap.2: MREMAP_DONTUNMAP to reflect to supported mappings

In reply to Brian Geffon

mremap(2) now supports MREMAP_DONTUNMAP with mapping types other than private anonymous.

---
 man2/mremap.2 | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)
man2/mremap.2 +2 -11
index 3ed0c0c0a..72acbc111 100644
@@ -118,16 +118,6 @@ This flag, which must be used in conjunction with
 remaps a mapping to a new address but does not unmap the mapping at
 .IR old_address .
 .IP
-The
-.B MREMAP_DONTUNMAP
-flag can be used only with private anonymous mappings
-(see the description of
-.BR MAP_PRIVATE
-and
-.BR MAP_ANONYMOUS
-in
-.BR mmap (2)).
-.IP
 After completion,
 any access to the range specified by
 .IR old_address
@@ -227,7 +217,8 @@ was specified, but one or more pages in the range specified by
 .IR old_address
 and
 .IR old_size
-were not private anonymous;
+were part of a special mapping or the mapping is one that
+does not support merging or expanding;
 .IP *
 .B MREMAP_DONTUNMAP
 was specified and
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
#30 lore raw

Re: [PATCH] mremap.2: MREMAP_DONTUNMAP to reflect to supported mappings

Hello Brian,

Is this already merged in Linux? I guess not, as I've seen a patch of yous for the kernel, right?

Thanks,

Alex

Quoted 40 lines · Brian Geffon

mremap(2) now supports MREMAP_DONTUNMAP with mapping types other than private anonymous.

---
  man2/mremap.2 | 13 ++-----------
  1 file changed, 2 insertions(+), 11 deletions(-)
man2/mremap.2 +2 -11
index 3ed0c0c0a..72acbc111 100644
@@ -118,16 +118,6 @@ This flag, which must be used in conjunction with
  remaps a mapping to a new address but does not unmap the mapping at
  .IR old_address .
  .IP
-The
-.B MREMAP_DONTUNMAP
-flag can be used only with private anonymous mappings
-(see the description of
-.BR MAP_PRIVATE
-and
-.BR MAP_ANONYMOUS
-in
-.BR mmap (2)).
-.IP
  After completion,
  any access to the range specified by
  .IR old_address
@@ -227,7 +217,8 @@ was specified, but one or more pages in the range specified by
  .IR old_address
  and
  .IR old_size
-were not private anonymous;
+were part of a special mapping or the mapping is one that
+does not support merging or expanding;
  .IP *
  .B MREMAP_DONTUNMAP
  was specified and
Signature · 3 lines
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/
#31 lore raw

Re: [PATCH] mremap.2: MREMAP_DONTUNMAP to reflect to supported mappings

Hi Alex,
It has not landed yet, it's currently in Andrew's mm tree. I can reach out again when it makes it into Linus' tree.

Brian

Quoted 56 lines · Alejandro Colomar (man-pages)

Hello Brian,

Is this already merged in Linux? I guess not, as I've seen a patch of yous for the kernel, right?

Thanks,

Alex

Quoted 31 lines · Brian Geffon

mremap(2) now supports MREMAP_DONTUNMAP with mapping types other than private anonymous.

---
  man2/mremap.2 | 13 ++-----------
  1 file changed, 2 insertions(+), 11 deletions(-)
man2/mremap.2 +0 -10
index 3ed0c0c0a..72acbc111 100644
@@ -118,16 +118,6 @@ This flag, which must be used in conjunction with
  remaps a mapping to a new address but does not unmap the mapping at
  .IR old_address .
  .IP
-The
-.B MREMAP_DONTUNMAP
-flag can be used only with private anonymous mappings
-(see the description of
-.BR MAP_PRIVATE
-and
-.BR MAP_ANONYMOUS
-in
-.BR mmap (2)).
-.IP
  After completion,
  any access to the range specified by
  .IR old_address

man2/mremap.2 was specified, but one or more pages in the range specified by · line 224

Quoted 9 lines
man2/mremap.2 +2 -1
@@ -227,7 +217,8 @@ was specified, but one or more pages in the range specified by
  .IR old_address
  and
  .IR old_size
-were not private anonymous;
+were part of a special mapping or the mapping is one that
+does not support merging or expanding;
  .IP *
  .B MREMAP_DONTUNMAP
  was specified and

--
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/

#32 lore raw

Re: [PATCH] mremap.2: MREMAP_DONTUNMAP to reflect to supported mappings

Quoted 5 lines · Brian Geffon

Hi Alex,
It has not landed yet, it's currently in Andrew's mm tree. I can reach out again when it makes it into Linus' tree.

Quoted 10 lines

Brian

Quoted 5 lines · Alejandro Colomar (man-pages)

Hello Brian,

Is this already merged in Linux? I guess not, as I've seen a patch of yous for the kernel, right?

Hi Alex,
This change has been in since 5.13 as commit a4609387859f0281951f5e476d9f76d7fb9ab321, I apologize I forgot to update you about the man page change after it landed.

Brian

Quoted 51 lines
Quoted 51 lines

Thanks,

Alex

Quoted 31 lines · Brian Geffon

mremap(2) now supports MREMAP_DONTUNMAP with mapping types other than private anonymous.

---
  man2/mremap.2 | 13 ++-----------
  1 file changed, 2 insertions(+), 11 deletions(-)
man2/mremap.2 +0 -10
index 3ed0c0c0a..72acbc111 100644
@@ -118,16 +118,6 @@ This flag, which must be used in conjunction with
  remaps a mapping to a new address but does not unmap the mapping at
  .IR old_address .
  .IP
-The
-.B MREMAP_DONTUNMAP
-flag can be used only with private anonymous mappings
-(see the description of
-.BR MAP_PRIVATE
-and
-.BR MAP_ANONYMOUS
-in
-.BR mmap (2)).
-.IP
  After completion,
  any access to the range specified by
  .IR old_address

man2/mremap.2 was specified, but one or more pages in the range specified by · line 224

Quoted 9 lines
man2/mremap.2 +2 -1
@@ -227,7 +217,8 @@ was specified, but one or more pages in the range specified by
  .IR old_address
  and
  .IR old_size
-were not private anonymous;
+were part of a special mapping or the mapping is one that
+does not support merging or expanding;
  .IP *
  .B MREMAP_DONTUNMAP
  was specified and

--
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/

Reviewed-by on v5 1/3
#33 lore raw

Re: [PATCH v5 1/3] mm: Extend MREMAP_DONTUNMAP to non-anonymous mappings

In reply to Brian Geffon
Quoted 10 lines · Brian Geffon

Currently MREMAP_DONTUNMAP only accepts private anonymous mappings. This restriction was placed initially for simplicity and not because there exists a technical reason to do so.

This change will widen the support to include any mappings which are not VM_DONTEXPAND or VM_PFNMAP. The primary use case is to support MREMAP_DONTUNMAP on mappings which may have been created from a memfd. This change will result in mremap(MREMAP_DONTUNMAP) returning -EINVAL if VM_DONTEXPAND or VM_PFNMAP mappings are specified.

Quoted 10 lines

Lokesh Gidra who works on the Android JVM, provided an explanation of how such a feature will improve Android JVM garbage collection:
"Android is developing a new garbage collector (GC), based on userfaultfd. The garbage collector will use userfaultfd (uffd) on the java heap during compaction. On accessing any uncompacted page, the application threads will find it missing, at which point the thread will create the compacted page and then use UFFDIO_COPY ioctl to get it mapped and then resume execution. Before starting this compaction, in a stop-the-world pause the heap will be mremap(MREMAP_DONTUNMAP) so that the java heap is ready to receive UFFD_EVENT_PAGEFAULT events after resuming execution.

Pretty interesting idea :-)

Quoted 15 lines

To speedup mremap operations, pagetable movement was optimized by moving PUD entries instead of PTE entries [1]. It was necessary as mremap of even modest sized memory ranges also took several milliseconds, and stopping the application for that long isn't acceptable in response-time sensitive cases.

With UFFDIO_CONTINUE feature [2], it will be even more efficient to implement this GC, particularly the 'non-moveable' portions of the heap. It will also help in reducing the need to copy (UFFDIO_COPY) the pages. However, for this to work, the java heap has to be on a 'shared' vma. Currently MREMAP_DONTUNMAP only supports private anonymous mappings, this patch will enable using UFFDIO_CONTINUE for the new userfaultfd-based heap compaction."

Quoted 6 lines

[1] https://lore.kernel.org/linux-mm/20201215030730.NC3CU98e4%[email protected]/
[2] https://lore.kernel.org/linux-mm/[email protected]/

Thanks,
Dmitry