Patch series "mm/mremap: permit mremap() move of multiple VMAs".

Historically we've made it a uAPI requirement that mremap() may only operate on a single VMA at a time.

For instances where VMAs need to be resized, this makes sense, as it becomes very difficult to determine what a user actually wants should they indicate a desire to expand or shrink the size of multiple VMAs (truncate? Adjust sizes individually? Some other strategy?).

However, in instances where a user is moving VMAs, it is restrictive to disallow this.

This is especially the case when anonymous mapping remap may or may not be mergeable depending on whether VMAs have or have not been faulted due to anon_vma assignment and folio index alignment with vma->vm_pgoff.

Often this can result in surprising impact where a moved region is faulted, then moved back and a user fails to observe a merge from otherwise compatible, adjacent VMAs.

This change allows such cases to work without the user having to be cognizant of whether a prior mremap() move or other VMA operations has resulted in VMA fragmentation.

In order to do this, this series performs a large amount of refactoring, most pertinently - grouping sanity checks together, separately those that check input parameters and those relating to VMAs.

We also simplify the post-mmap lock drop processing for uffd and mlock()'d VMAs.

With this done, we can then fairly straightforwardly implement this functionality.

This works exclusively for mremap() invocations which specify MREMAP_FIXED. It is not compatible with VMAs which use userfaultfd, as the notification of the userland fault handler would require us to drop the mmap lock.

The input and output addresses ranges must not overlap. We carefully account for moves which would result in VMA merges or would otherwise result in VMA iterator invalidation.

This patch (of 10):

We const-ify the vrm flags parameter to indicate this will never change.

We rename resize_is_valid() to remap_is_valid(), as this function does not only apply to cases where we resize, so it's simply confusing to refer to that here.

We remove the BUG() from mremap_at(), as we should not BUG() unless we are certain it'll result in system instability.

We rename vrm_charge() to vrm_calc_charge() to make it clear this simply calculates the charged number of pages rather than actually adjusting any state.

We update the comment for vrm_implies_new_addr() to explain that MREMAP_DONTUNMAP does not require a set address, but will always be moved.

Additionally consistently use 'res' rather than 'ret' for result values.

No functional change intended.

Signed-off-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com Reviewed-by: Vlastimil Babka vbabka@suse.cz --- mm/mremap.c | 55 +++++++++++++++++++++++++++++++---------------------- 1 file changed, 32 insertions(+), 23 deletions(-)

diff --git a/mm/mremap.c b/mm/mremap.c index 36585041c760..1815095c4bca 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -52,7 +52,7 @@ struct vma_remap_struct { unsigned long addr; /* User-specified address from which we remap. */ unsigned long old_len; /* Length of range being remapped. */ unsigned long new_len; /* Desired new length of mapping. */ - unsigned long flags; /* user-specified MREMAP_* flags. */ + const unsigned long flags; /* user-specified MREMAP_* flags. */ unsigned long new_addr; /* Optionally, desired new address. */

/* uffd state. */ @@ -911,7 +911,11 @@ static bool vrm_overlaps(struct vma_remap_struct *vrm) return false; }

-/* Do the mremap() flags require that the new_addr parameter be specified? */ +/* + * Will a new address definitely be assigned? This either if the user specifies + * it via MREMAP_FIXED, or if MREMAP_DONTUNMAP is used, indicating we will + * always detemrine a target address. + */ static bool vrm_implies_new_addr(struct vma_remap_struct *vrm) { return vrm->flags & (MREMAP_FIXED | MREMAP_DONTUNMAP); @@ -957,7 +961,7 @@ static unsigned long vrm_set_new_addr(struct vma_remap_struct *vrm) * * Returns true on success, false if insufficient memory to charge. */ -static bool vrm_charge(struct vma_remap_struct *vrm) +static bool vrm_calc_charge(struct vma_remap_struct *vrm) { unsigned long charged;

@@ -1262,8 +1266,11 @@ static unsigned long move_vma(struct vma_remap_struct *vrm) if (err) return err;

- /* If accounted, charge the number of bytes the operation will use. */ - if (!vrm_charge(vrm)) + /* + * If accounted, determine the number of bytes the operation will + * charge. + */ + if (!vrm_calc_charge(vrm)) return -ENOMEM;

/* We don't want racing faults. */ @@ -1302,12 +1309,12 @@ static unsigned long move_vma(struct vma_remap_struct *vrm) }

/* - * resize_is_valid() - Ensure the vma can be resized to the new length at the give - * address. + * remap_is_valid() - Ensure the VMA can be moved or resized to the new length, + * at the given address. * * Return 0 on success, error otherwise. */ -static int resize_is_valid(struct vma_remap_struct *vrm) +static int remap_is_valid(struct vma_remap_struct *vrm) { struct mm_struct *mm = current->mm; struct vm_area_struct *vma = vrm->vma; @@ -1446,7 +1453,7 @@ static unsigned long mremap_to(struct vma_remap_struct *vrm) vrm->old_len = vrm->new_len; }

- err = resize_is_valid(vrm); + err = remap_is_valid(vrm); if (err) return err;

@@ -1571,7 +1578,7 @@ static unsigned long expand_vma_in_place(struct vma_remap_struct *vrm) struct vm_area_struct *vma = vrm->vma; VMA_ITERATOR(vmi, mm, vma->vm_end);

- if (!vrm_charge(vrm)) + if (!vrm_calc_charge(vrm)) return -ENOMEM;

/* @@ -1632,7 +1639,7 @@ static unsigned long expand_vma(struct vma_remap_struct *vrm) unsigned long err; unsigned long addr = vrm->addr;

- err = resize_is_valid(vrm); + err = remap_is_valid(vrm); if (err) return err;

@@ -1705,18 +1712,20 @@ static unsigned long mremap_at(struct vma_remap_struct *vrm) return expand_vma(vrm); }

- BUG(); + /* Should not be possible. */ + WARN_ON_ONCE(1); + return -EINVAL; }

static unsigned long do_mremap(struct vma_remap_struct *vrm) { struct mm_struct *mm = current->mm; struct vm_area_struct *vma; - unsigned long ret; + unsigned long res;

- ret = check_mremap_params(vrm); - if (ret) - return ret; + res = check_mremap_params(vrm); + if (res) + return res;

vrm->old_len = PAGE_ALIGN(vrm->old_len); vrm->new_len = PAGE_ALIGN(vrm->new_len); @@ -1728,41 +1737,41 @@ static unsigned long do_mremap(struct vma_remap_struct *vrm)

vma = vrm->vma = vma_lookup(mm, vrm->addr); if (!vma) { - ret = -EFAULT; + res = -EFAULT; goto out; }

/* If mseal()'d, mremap() is prohibited. */ if (!can_modify_vma(vma)) { - ret = -EPERM; + res = -EPERM; goto out; }

/* Align to hugetlb page size, if required. */ if (is_vm_hugetlb_page(vma) && !align_hugetlb(vrm)) { - ret = -EINVAL; + res = -EINVAL; goto out; }

vrm->remap_type = vrm_remap_type(vrm);

/* Actually execute mremap. */ - ret = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm); + res = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm);

out: if (vrm->mmap_locked) { mmap_write_unlock(mm); vrm->mmap_locked = false;

- if (!offset_in_page(ret) && vrm->mlocked && vrm->new_len > vrm->old_len) + if (!offset_in_page(res) && vrm->mlocked && vrm->new_len > vrm->old_len) mm_populate(vrm->new_addr + vrm->old_len, vrm->delta); }

userfaultfd_unmap_complete(mm, vrm->uf_unmap_early); - mremap_userfaultfd_complete(vrm->uf, vrm->addr, ret, vrm->old_len); + mremap_userfaultfd_complete(vrm->uf, vrm->addr, res, vrm->old_len); userfaultfd_unmap_complete(mm, vrm->uf_unmap);

- return ret; + return res; }

/*

Rather than lumping everything together in do_mremap(), add a new helper function, check_prep_vma(), to do the work relating to each VMA.

This further lays groundwork for subsequent patches which will allow for batched VMA mremap().

Additionally, if we set vrm->new_addr == vrm->addr when prepping the VMA, this avoids us needing to do so in the expand VMA mlocked case.

No functional change intended.

Signed-off-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com Reviewed-by: Vlastimil Babka vbabka@suse.cz --- mm/mremap.c | 58 ++++++++++++++++++++++++++--------------------------- 1 file changed, 28 insertions(+), 30 deletions(-)

diff --git a/mm/mremap.c b/mm/mremap.c index a00da0288c37..d57645573e0d 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -1636,7 +1636,6 @@ static bool align_hugetlb(struct vma_remap_struct *vrm) static unsigned long expand_vma(struct vma_remap_struct *vrm) { unsigned long err; - unsigned long addr = vrm->addr;

err = remap_is_valid(vrm); if (err) @@ -1651,16 +1650,8 @@ static unsigned long expand_vma(struct vma_remap_struct *vrm) if (err) return err;

- /* - * We want to populate the newly expanded portion of the VMA to - * satisfy the expectation that mlock()'ing a VMA maintains all - * of its pages in memory. - */ - if (vrm->mlocked) - vrm->new_addr = addr; - /* OK we're done! */ - return addr; + return vrm->addr; }

/* @@ -1716,10 +1707,33 @@ static unsigned long mremap_at(struct vma_remap_struct *vrm) return -EINVAL; }

+static int check_prep_vma(struct vma_remap_struct *vrm) +{ + struct vm_area_struct *vma = vrm->vma; + + if (!vma) + return -EFAULT; + + /* If mseal()'d, mremap() is prohibited. */ + if (!can_modify_vma(vma)) + return -EPERM; + + /* Align to hugetlb page size, if required. */ + if (is_vm_hugetlb_page(vma) && !align_hugetlb(vrm)) + return -EINVAL; + + vrm_set_delta(vrm); + vrm->remap_type = vrm_remap_type(vrm); + /* For convenience, we set new_addr even if VMA won't move. */ + if (!vrm_implies_new_addr(vrm)) + vrm->new_addr = vrm->addr; + + return 0; +} + static unsigned long do_mremap(struct vma_remap_struct *vrm) { struct mm_struct *mm = current->mm; - struct vm_area_struct *vma; unsigned long res;

vrm->old_len = PAGE_ALIGN(vrm->old_len); @@ -1733,26 +1747,10 @@ static unsigned long do_mremap(struct vma_remap_struct *vrm) return -EINTR; vrm->mmap_locked = true;

- vma = vrm->vma = vma_lookup(mm, vrm->addr); - if (!vma) { - res = -EFAULT; - goto out; - } - - /* If mseal()'d, mremap() is prohibited. */ - if (!can_modify_vma(vma)) { - res = -EPERM; - goto out; - } - - /* Align to hugetlb page size, if required. */ - if (is_vm_hugetlb_page(vma) && !align_hugetlb(vrm)) { - res = -EINVAL; + vrm->vma = vma_lookup(current->mm, vrm->addr); + res = check_prep_vma(vrm); + if (res) goto out; - } - - vrm_set_delta(vrm); - vrm->remap_type = vrm_remap_type(vrm);

/* Actually execute mremap. */ res = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm);

Right now it appears that the code is relying upon the returned destination address having bits outside PAGE_MASK to indicate whether an error value is specified, and decrementing the increased refcount on the uffd ctx if so.

This is not a safe means of determining an error value, so instead, be specific. It makes far more sense to do so in a dedicated error path, so add mremap_userfaultfd_fail() for this purpose and use this when an error arises.

A vm_userfaultfd_ctx is not established until we are at the point where mremap_userfaultfd_prep() is invoked in copy_vma_and_data(), so this is a no-op until this happens.

That is - uffd remap notification only occurs if the VMA is actually moved - at which point a UFFD_EVENT_REMAP event is raised.

No errors can occur after this point currently, though it's certainly not guaranteed this will always remain the case, and we mustn't rely on this.

However, the reason for needing to handle this case is that, when an error arises on a VMA move at the point of adjusting page tables, we revert this operation, and propagate the error.

At this point, it is not correct to raise a uffd remap event, and we must handle it.

This refactoring makes it abundantly clear what we are doing.

We assume vrm->new_addr is always valid, which a prior change made the case even for mremap() invocations which don't move the VMA, however given no uffd context would be set up in this case it's immaterial to this change anyway.

No functional change intended.

Signed-off-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com Reviewed-by: Vlastimil Babka vbabka@suse.cz --- fs/userfaultfd.c | 15 ++++++++++----- include/linux/userfaultfd_k.h | 5 +++++ mm/mremap.c | 16 ++++++++++++---- 3 files changed, 27 insertions(+), 9 deletions(-)

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c index 2a644aa1a510..54c6cc7fe9c6 100644 --- a/fs/userfaultfd.c +++ b/fs/userfaultfd.c @@ -750,11 +750,6 @@ void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *vm_ctx, if (!ctx) return;

- if (to & ~PAGE_MASK) { - userfaultfd_ctx_put(ctx); - return; - } - msg_init(&ewq.msg);

ewq.msg.event = UFFD_EVENT_REMAP; @@ -765,6 +760,16 @@ void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *vm_ctx, userfaultfd_event_wait_completion(ctx, &ewq); }

+void mremap_userfaultfd_fail(struct vm_userfaultfd_ctx *vm_ctx) +{ + struct userfaultfd_ctx *ctx = vm_ctx->ctx; + + if (!ctx) + return; + + userfaultfd_ctx_put(ctx); +} + bool userfaultfd_remove(struct vm_area_struct *vma, unsigned long start, unsigned long end) { diff --git a/include/linux/userfaultfd_k.h b/include/linux/userfaultfd_k.h index df85330bcfa6..c0e716aec26a 100644 --- a/include/linux/userfaultfd_k.h +++ b/include/linux/userfaultfd_k.h @@ -259,6 +259,7 @@ extern void mremap_userfaultfd_prep(struct vm_area_struct *, extern void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *, unsigned long from, unsigned long to, unsigned long len); +void mremap_userfaultfd_fail(struct vm_userfaultfd_ctx *);

extern bool userfaultfd_remove(struct vm_area_struct *vma, unsigned long start, @@ -371,6 +372,10 @@ static inline void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *ctx, { }

+static inline void mremap_userfaultfd_fail(struct vm_userfaultfd_ctx *ctx) +{ +} + static inline bool userfaultfd_remove(struct vm_area_struct *vma, unsigned long start, unsigned long end) diff --git a/mm/mremap.c b/mm/mremap.c index 87cab223f2bb..b2ee95182b36 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -1731,12 +1731,17 @@ static int check_prep_vma(struct vma_remap_struct *vrm) return 0; }

-static void notify_uffd(struct vma_remap_struct *vrm, unsigned long to) +static void notify_uffd(struct vma_remap_struct *vrm, bool failed) { struct mm_struct *mm = current->mm;

+ /* Regardless of success/failure, we always notify of any unmaps. */ userfaultfd_unmap_complete(mm, vrm->uf_unmap_early); - mremap_userfaultfd_complete(vrm->uf, vrm->addr, to, vrm->old_len); + if (failed) + mremap_userfaultfd_fail(vrm->uf); + else + mremap_userfaultfd_complete(vrm->uf, vrm->addr, + vrm->new_addr, vrm->old_len); userfaultfd_unmap_complete(mm, vrm->uf_unmap); }

@@ -1744,6 +1749,7 @@ static unsigned long do_mremap(struct vma_remap_struct *vrm) { struct mm_struct *mm = current->mm; unsigned long res; + bool failed;

vrm->old_len = PAGE_ALIGN(vrm->old_len); vrm->new_len = PAGE_ALIGN(vrm->new_len); @@ -1765,13 +1771,15 @@ static unsigned long do_mremap(struct vma_remap_struct *vrm) res = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm);

out: + failed = IS_ERR_VALUE(res); + if (vrm->mmap_locked) mmap_write_unlock(mm);

- if (!IS_ERR_VALUE(res) && vrm->mlocked && vrm->new_len > vrm->old_len) + if (!failed && vrm->mlocked && vrm->new_len > vrm->old_len) mm_populate(vrm->new_addr + vrm->old_len, vrm->delta);

- notify_uffd(vrm, res); + notify_uffd(vrm, failed); return res; }

Group parameter check logic together, moving check_mremap_params() next to it.

This puts all such checks into a single place, and invokes them early so we can simply bail out as soon as we are aware that a condition is not met.

No functional change intended.

Signed-off-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com Reviewed-by: Vlastimil Babka vbabka@suse.cz --- mm/mremap.c | 273 +++++++++++++++++++++++++--------------------------- 1 file changed, 131 insertions(+), 142 deletions(-)

diff --git a/mm/mremap.c b/mm/mremap.c index 116203766ce0..3f8daa3314f0 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -1308,64 +1308,6 @@ static unsigned long move_vma(struct vma_remap_struct *vrm) return err ? (unsigned long)err : vrm->new_addr; }

-/* - * remap_is_valid() - Ensure the VMA can be moved or resized to the new length, - * at the given address. - * - * Return 0 on success, error otherwise. - */ -static int remap_is_valid(struct vma_remap_struct *vrm) -{ - struct mm_struct *mm = current->mm; - struct vm_area_struct *vma = vrm->vma; - unsigned long addr = vrm->addr; - unsigned long old_len = vrm->old_len; - unsigned long new_len = vrm->new_len; - unsigned long pgoff; - - /* - * !old_len is a special case where an attempt is made to 'duplicate' - * a mapping. This makes no sense for private mappings as it will - * instead create a fresh/new mapping unrelated to the original. This - * is contrary to the basic idea of mremap which creates new mappings - * based on the original. There are no known use cases for this - * behavior. As a result, fail such attempts. - */ - if (!old_len && !(vma->vm_flags & (VM_SHARED | VM_MAYSHARE))) { - pr_warn_once("%s (%d): attempted to duplicate a private mapping with mremap. This is not supported.\n", - current->comm, current->pid); - return -EINVAL; - } - - if ((vrm->flags & MREMAP_DONTUNMAP) && - (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP))) - return -EINVAL; - - /* We can't remap across vm area boundaries */ - if (old_len > vma->vm_end - addr) - return -EFAULT; - - if (new_len <= old_len) - return 0; - - /* Need to be careful about a growing mapping */ - pgoff = (addr - vma->vm_start) >> PAGE_SHIFT; - pgoff += vma->vm_pgoff; - if (pgoff + (new_len >> PAGE_SHIFT) < pgoff) - return -EINVAL; - - if (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)) - return -EFAULT; - - if (!mlock_future_ok(mm, vma->vm_flags, vrm->delta)) - return -EAGAIN; - - if (!may_expand_vm(mm, vma->vm_flags, vrm->delta >> PAGE_SHIFT)) - return -ENOMEM; - - return 0; -} - /* * The user has requested that the VMA be shrunk (i.e., old_len > new_len), so * execute this, optionally dropping the mmap lock when we do so. @@ -1492,77 +1434,6 @@ static bool vrm_can_expand_in_place(struct vma_remap_struct *vrm) return true; }

-/* - * Are the parameters passed to mremap() valid? If so return 0, otherwise return - * error. - */ -static unsigned long check_mremap_params(struct vma_remap_struct *vrm) - -{ - unsigned long addr = vrm->addr; - unsigned long flags = vrm->flags; - - /* Ensure no unexpected flag values. */ - if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE | MREMAP_DONTUNMAP)) - return -EINVAL; - - /* Start address must be page-aligned. */ - if (offset_in_page(addr)) - return -EINVAL; - - /* - * We allow a zero old-len as a special case - * for DOS-emu "duplicate shm area" thing. But - * a zero new-len is nonsensical. - */ - if (!vrm->new_len) - return -EINVAL; - - /* Is the new length or address silly? */ - if (vrm->new_len > TASK_SIZE || - vrm->new_addr > TASK_SIZE - vrm->new_len) - return -EINVAL; - - /* Remainder of checks are for cases with specific new_addr. */ - if (!vrm_implies_new_addr(vrm)) - return 0; - - /* The new address must be page-aligned. */ - if (offset_in_page(vrm->new_addr)) - return -EINVAL; - - /* A fixed address implies a move. */ - if (!(flags & MREMAP_MAYMOVE)) - return -EINVAL; - - /* MREMAP_DONTUNMAP does not allow resizing in the process. */ - if (flags & MREMAP_DONTUNMAP && vrm->old_len != vrm->new_len) - return -EINVAL; - - /* Target VMA must not overlap source VMA. */ - if (vrm_overlaps(vrm)) - return -EINVAL; - - /* - * move_vma() need us to stay 4 maps below the threshold, otherwise - * it will bail out at the very beginning. - * That is a problem if we have already unmaped the regions here - * (new_addr, and old_addr), because userspace will not know the - * state of the vma's after it gets -ENOMEM. - * So, to avoid such scenario we can pre-compute if the whole - * operation has high chances to success map-wise. - * Worst-scenario case is when both vma's (new_addr and old_addr) get - * split in 3 before unmapping it. - * That means 2 more maps (1 for each) to the ones we already hold. - * Check whether current map count plus 2 still leads us to 4 maps below - * the threshold, otherwise return -ENOMEM here to be more safe. - */ - if ((current->mm->map_count + 2) >= sysctl_max_map_count - 3) - return -ENOMEM; - - return 0; -} - /* * We know we can expand the VMA in-place by delta pages, so do so. * @@ -1714,9 +1585,26 @@ static bool vrm_will_map_new(struct vma_remap_struct *vrm) return false; }

+static void notify_uffd(struct vma_remap_struct *vrm, bool failed) +{ + struct mm_struct *mm = current->mm; + + /* Regardless of success/failure, we always notify of any unmaps. */ + userfaultfd_unmap_complete(mm, vrm->uf_unmap_early); + if (failed) + mremap_userfaultfd_fail(vrm->uf); + else + mremap_userfaultfd_complete(vrm->uf, vrm->addr, + vrm->new_addr, vrm->old_len); + userfaultfd_unmap_complete(mm, vrm->uf_unmap); +} + static int check_prep_vma(struct vma_remap_struct *vrm) { struct vm_area_struct *vma = vrm->vma; + struct mm_struct *mm = current->mm; + unsigned long addr = vrm->addr; + unsigned long old_len, new_len, pgoff;

if (!vma) return -EFAULT; @@ -1733,26 +1621,127 @@ static int check_prep_vma(struct vma_remap_struct *vrm) vrm->remap_type = vrm_remap_type(vrm); /* For convenience, we set new_addr even if VMA won't move. */ if (!vrm_implies_new_addr(vrm)) - vrm->new_addr = vrm->addr; + vrm->new_addr = addr; + + /* Below only meaningful if we expand or move a VMA. */ + if (!vrm_will_map_new(vrm)) + return 0;

- if (vrm_will_map_new(vrm)) - return remap_is_valid(vrm); + old_len = vrm->old_len; + new_len = vrm->new_len; + + /* + * !old_len is a special case where an attempt is made to 'duplicate' + * a mapping. This makes no sense for private mappings as it will + * instead create a fresh/new mapping unrelated to the original. This + * is contrary to the basic idea of mremap which creates new mappings + * based on the original. There are no known use cases for this + * behavior. As a result, fail such attempts. + */ + if (!old_len && !(vma->vm_flags & (VM_SHARED | VM_MAYSHARE))) { + pr_warn_once("%s (%d): attempted to duplicate a private mapping with mremap. This is not supported.\n", + current->comm, current->pid); + return -EINVAL; + } + + if ((vrm->flags & MREMAP_DONTUNMAP) && + (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP))) + return -EINVAL; + + /* We can't remap across vm area boundaries */ + if (old_len > vma->vm_end - addr) + return -EFAULT; + + if (new_len <= old_len) + return 0; + + /* Need to be careful about a growing mapping */ + pgoff = (addr - vma->vm_start) >> PAGE_SHIFT; + pgoff += vma->vm_pgoff; + if (pgoff + (new_len >> PAGE_SHIFT) < pgoff) + return -EINVAL; + + if (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)) + return -EFAULT; + + if (!mlock_future_ok(mm, vma->vm_flags, vrm->delta)) + return -EAGAIN; + + if (!may_expand_vm(mm, vma->vm_flags, vrm->delta >> PAGE_SHIFT)) + return -ENOMEM;

return 0; }

-static void notify_uffd(struct vma_remap_struct *vrm, bool failed) +/* + * Are the parameters passed to mremap() valid? If so return 0, otherwise return + * error. + */ +static unsigned long check_mremap_params(struct vma_remap_struct *vrm) + { - struct mm_struct *mm = current->mm; + unsigned long addr = vrm->addr; + unsigned long flags = vrm->flags;

- /* Regardless of success/failure, we always notify of any unmaps. */ - userfaultfd_unmap_complete(mm, vrm->uf_unmap_early); - if (failed) - mremap_userfaultfd_fail(vrm->uf); - else - mremap_userfaultfd_complete(vrm->uf, vrm->addr, - vrm->new_addr, vrm->old_len); - userfaultfd_unmap_complete(mm, vrm->uf_unmap); + /* Ensure no unexpected flag values. */ + if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE | MREMAP_DONTUNMAP)) + return -EINVAL; + + /* Start address must be page-aligned. */ + if (offset_in_page(addr)) + return -EINVAL; + + /* + * We allow a zero old-len as a special case + * for DOS-emu "duplicate shm area" thing. But + * a zero new-len is nonsensical. + */ + if (!vrm->new_len) + return -EINVAL; + + /* Is the new length or address silly? */ + if (vrm->new_len > TASK_SIZE || + vrm->new_addr > TASK_SIZE - vrm->new_len) + return -EINVAL; + + /* Remainder of checks are for cases with specific new_addr. */ + if (!vrm_implies_new_addr(vrm)) + return 0; + + /* The new address must be page-aligned. */ + if (offset_in_page(vrm->new_addr)) + return -EINVAL; + + /* A fixed address implies a move. */ + if (!(flags & MREMAP_MAYMOVE)) + return -EINVAL; + + /* MREMAP_DONTUNMAP does not allow resizing in the process. */ + if (flags & MREMAP_DONTUNMAP && vrm->old_len != vrm->new_len) + return -EINVAL; + + /* Target VMA must not overlap source VMA. */ + if (vrm_overlaps(vrm)) + return -EINVAL; + + /* + * move_vma() need us to stay 4 maps below the threshold, otherwise + * it will bail out at the very beginning. + * That is a problem if we have already unmaped the regions here + * (new_addr, and old_addr), because userspace will not know the + * state of the vma's after it gets -ENOMEM. + * So, to avoid such scenario we can pre-compute if the whole + * operation has high chances to success map-wise. + * Worst-scenario case is when both vma's (new_addr and old_addr) get + * split in 3 before unmapping it. + * That means 2 more maps (1 for each) to the ones we already hold. + * Check whether current map count plus 2 still leads us to 4 maps below + * the threshold, otherwise return -ENOMEM here to be more safe. + */ + if ((current->mm->map_count + 2) >= sysctl_max_map_count - 3) + return -ENOMEM; + + return 0; }

static unsigned long do_mremap(struct vma_remap_struct *vrm)

Historically we've made it a uAPI requirement that mremap() may only operate on a single VMA at a time.

For instances where VMAs need to be resized, this makes sense, as it becomes very difficult to determine what a user actually wants should they indicate a desire to expand or shrink the size of multiple VMAs (truncate? Adjust sizes individually? Some other strategy?).

However, in instances where a user is moving VMAs, it is restrictive to disallow this.

This is especially the case when anonymous mapping remap may or may not be mergeable depending on whether VMAs have or have not been faulted due to anon_vma assignment and folio index alignment with vma->vm_pgoff.

Often this can result in surprising impact where a moved region is faulted, then moved back and a user fails to observe a merge from otherwise compatible, adjacent VMAs.

This change allows such cases to work without the user having to be cognizant of whether a prior mremap() move or other VMA operations has resulted in VMA fragmentation.

Having refactored mremap code to aggregate per-VMA and parameter checks, we are now in a position to permit this kind of move.

We do so by detecting if this is a move-only operation up-front, and then utilising a separate code path via remap_move() rather than the ordinary single-VMA path.

There are two tasks that occur outside of the mmap write lock - userfaultfd notification and population of unmapped regions of expanded VMAs should the VMA be mlock()'d.

The latter doesn't apply, as this is logic for a move only and thus no expansion can take place. In the former case, we explicitly disallow multi-VMA operations on uffd-armed VMAs.

The mmap lock is never dropped in the move-only case, this only occurs on a VMA shrink.

We take care to handle cases where a VMA merge has occurred, by resetting the VMA iterator in such instances.

We needn't worry about self-merges, as in those cases we would, by definition, not be spanning multiple VMAs. The overlapping range test is performed on the whole range so specifically disallows this.

Signed-off-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com --- mm/mremap.c | 132 +++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 125 insertions(+), 7 deletions(-)

diff --git a/mm/mremap.c b/mm/mremap.c index 8cb08ccea6ad..1c88ed7eae0a 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -69,6 +69,8 @@ struct vma_remap_struct { enum mremap_type remap_type; /* expand, shrink, etc. */ bool mmap_locked; /* Is mm currently write-locked? */ unsigned long charged; /* If VM_ACCOUNT, # pages to account. */ + bool multi_vma; /* Is >1 VMA being moved? */ + bool vmi_needs_reset; /* Was the VMA iterator invalidated? */ };

static pud_t *get_old_pud(struct mm_struct *mm, unsigned long addr) @@ -1188,6 +1190,9 @@ static int copy_vma_and_data(struct vma_remap_struct *vrm, *new_vma_ptr = NULL; return -ENOMEM; } + if (vma != vrm->vma) + vrm->vmi_needs_reset = true; + vrm->vma = vma; pmc.old = vma; pmc.new = new_vma; @@ -1583,6 +1588,18 @@ static bool vrm_will_map_new(struct vma_remap_struct *vrm) return false; }

+/* Does this remap ONLY move mappings? */ +static bool vrm_move_only(struct vma_remap_struct *vrm) +{ + if (!vrm_implies_new_addr(vrm)) + return false; + + if (vrm->old_len != vrm->new_len) + return false; + + return true; +} + static void notify_uffd(struct vma_remap_struct *vrm, bool failed) { struct mm_struct *mm = current->mm; @@ -1646,10 +1663,29 @@ static int check_prep_vma(struct vma_remap_struct *vrm) (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP))) return -EINVAL;

- /* We can't remap across vm area boundaries */ + /* + * We can't remap across the end of VMAs, as another VMA may be + * adjacent: + * + * addr vma->vm_end + * |-----.----------| + * | . | + * |-----.----------| + * .<--------->xxx> + * old_len + * + * We also require that vma->vm_start <= addr < vma->vm_end. + */ if (old_len > vma->vm_end - addr) return -EFAULT;

+ /* + * We can't support moving multiple uffd VMAs as notify requires mmap + * lock to be dropped. + */ + if (vrm->multi_vma && userfaultfd_armed(vma)) + return -EINVAL; + if (new_len <= old_len) return 0;

@@ -1746,6 +1782,84 @@ static unsigned long check_mremap_params(struct vma_remap_struct *vrm) return 0; }

+static unsigned long remap_move(struct vma_remap_struct *vrm) +{ + struct vm_area_struct *vma; + unsigned long start = vrm->addr; + unsigned long end = vrm->addr + vrm->old_len; + unsigned long new_addr = vrm->new_addr; + unsigned long prev_addr = start; + VMA_ITERATOR(vmi, current->mm, start); + unsigned long res = -EFAULT; + unsigned long last_end; + + /* + * When moving VMAs we allow for batched moves across multiple VMAs, + * with all VMAs in the input range [addr, addr + old_len) being moved + * (and split as necessary). + */ + for_each_vma_range(vmi, vma, end) { + /* Account for start, end not aligned with VMA start, end. */ + unsigned long addr = max(vma->vm_start, start); + unsigned long len = min(end, vma->vm_end) - addr; + unsigned long offset, res_vma; + + /* Merged with self, move on. */ + if (vrm->multi_vma && prev_addr == addr) + continue; + + /* + * To sensibly move multiple VMAs, accounting for the fact that + * get_unmapped_area() may align even MAP_FIXED moves, we simply + * attempt to move such that the gaps between source VMAs remain + * consistent in destination VMAs, e.g.: + * + * X Y X Y + * <---> <-> <---> <-> + * |-------| |-----| |-----| |-------| |-----| |-----| + * | A | | B | | C | ---> | A' | | B' | | C' | + * |-------| |-----| |-----| |-------| |-----| |-----| + * new_addr + * + * Now, new_addr may be altered even with MREMAP_FIXED set, due + * to e.g. alignment changes from get_unmapped_area(). + * + * So we map B' at A'->vm_end + X, and C' at B'->vm_end + Y. + */ + offset = vrm->multi_vma ? vma->vm_start - last_end : 0; + last_end = vma->vm_end; + + vrm->vma = vma; + vrm->addr = addr; + vrm->new_addr = new_addr + offset; + vrm->old_len = vrm->new_len = len; + + res_vma = check_prep_vma(vrm); + if (!res_vma) + res_vma = mremap_to(vrm); + if (IS_ERR_VALUE(res_vma)) + return res_vma; + + /* mmap lock is only dropped on shrink. */ + VM_WARN_ON_ONCE(!vrm->mmap_locked); + /* This is a move, no expand should occur. */ + VM_WARN_ON_ONCE(vrm->populate_expand); + + if (!vrm->multi_vma) + res = res_vma; + + if (vrm->vmi_needs_reset) { + vma_iter_reset(&vmi); + vrm->vmi_needs_reset = false; + } + vrm->multi_vma = true; + prev_addr = addr; + new_addr = res_vma + vrm->new_len; + } + + return res; +} + static unsigned long do_mremap(struct vma_remap_struct *vrm) { struct mm_struct *mm = current->mm; @@ -1763,13 +1877,17 @@ static unsigned long do_mremap(struct vma_remap_struct *vrm) return -EINTR; vrm->mmap_locked = true;

- vrm->vma = vma_lookup(current->mm, vrm->addr); - res = check_prep_vma(vrm); - if (res) - goto out; + if (vrm_move_only(vrm)) { + res = remap_move(vrm); + } else { + vrm->vma = vma_lookup(current->mm, vrm->addr); + res = check_prep_vma(vrm); + if (res) + goto out;

- /* Actually execute mremap. */ - res = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm); + /* Actually execute mremap. */ + res = vrm_implies_new_addr(vrm) ? mremap_to(vrm) : mremap_at(vrm); + }

out: failed = IS_ERR_VALUE(res);