| Message ID | 150846720244.24336.16885325309403883980.stgit@dwillia2-desk3.amr.corp.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, 2017-10-19 at 19:40 -0700, Dan Williams wrote: > get_user_pages() pins file backed memory pages for access by dma > devices. However, it only pins the memory pages not the page-to-file > offset association. If a file is truncated the pages are mapped out of > the file and dma may continue indefinitely into a page that is owned by > a device driver. This breaks coherency of the file vs dma, but the > assumption is that if userspace wants the file-space truncated it does > not matter what data is inbound from the device, it is not relevant > anymore. > > The assumptions of the truncate-page-cache model are broken by DAX where > the target DMA page *is* the filesystem block. Leaving the page pinned > for DMA, but truncating the file block out of the file, means that the > filesytem is free to reallocate a block under active DMA to another > file! > > Here are some possible options for fixing this situation ('truncate' and > 'fallocate(punch hole)' are synonymous below): > > 1/ Fail truncate while any file blocks might be under dma > > 2/ Block (sleep-wait) truncate while any file blocks might be under > dma > > 3/ Remap file blocks to a "lost+found"-like file-inode where > dma can continue and we might see what inbound data from DMA was > mapped out of the original file. Blocks in this file could be > freed back to the filesystem when dma eventually ends. > > 4/ Disable dax until option 3 or another long term solution has been > implemented. However, filesystem-dax is still marked experimental > for concerns like this. > > Option 1 will throw failures where userspace has never expected them > before, option 2 might hang the truncating process indefinitely, and > option 3 requires per filesystem enabling to remap blocks from one inode > to another. Option 2 is implemented in this patch for the DAX path with > the expectation that non-transient users of get_user_pages() (RDMA) are > disallowed from setting up dax mappings and that the potential delay > introduced to the truncate path is acceptable compared to the response > time of the page cache case. This can only be seen as a stop-gap until > we can solve the problem of safely sequestering unallocated filesystem > blocks under active dma. > FWIW, I like #3 a lot more than #2 here. I get that it's quite a bit more work though, so no objection to this as a stop-gap fix. > The solution introduces a new FL_ALLOCATED lease to pin the allocated > blocks in a dax file while dma might be accessing them. It behaves > identically to an FL_LAYOUT lease save for the fact that it is > immediately sheduled to be reaped, and that the only path that waits for > its removal is the truncate path. We can not reuse FL_LAYOUT directly > since that would deadlock in the case where userspace did a direct-I/O > operation with a target buffer backed by an mmap range of the same file. > > Credit / inspiration for option 3 goes to Dave Hansen, who proposed > something similar as an alternative way to solve the problem that > MAP_DIRECT was trying to solve. > > Cc: Jan Kara <jack@suse.cz> > Cc: Jeff Moyer <jmoyer@redhat.com> > Cc: Dave Chinner <david@fromorbit.com> > Cc: Matthew Wilcox <mawilcox@microsoft.com> > Cc: Alexander Viro <viro@zeniv.linux.org.uk> > Cc: "Darrick J. Wong" <darrick.wong@oracle.com> > Cc: Ross Zwisler <ross.zwisler@linux.intel.com> > Cc: Jeff Layton <jlayton@poochiereds.net> > Cc: "J. Bruce Fields" <bfields@fieldses.org> > Cc: Dave Hansen <dave.hansen@linux.intel.com> > Reported-by: Christoph Hellwig <hch@lst.de> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> > --- > fs/Kconfig | 1 > fs/dax.c | 188 +++++++++++++++++++++++++++++++++++++++++++++++++++ > fs/locks.c | 17 ++++- > include/linux/dax.h | 23 ++++++ > include/linux/fs.h | 22 +++++- > mm/gup.c | 27 ++++++- > 6 files changed, 268 insertions(+), 10 deletions(-) > > diff --git a/fs/Kconfig b/fs/Kconfig > index 7aee6d699fd6..a7b31a96a753 100644 > --- a/fs/Kconfig > +++ b/fs/Kconfig > @@ -37,6 +37,7 @@ source "fs/f2fs/Kconfig" > config FS_DAX > bool "Direct Access (DAX) support" > depends on MMU > + depends on FILE_LOCKING > depends on !(ARM || MIPS || SPARC) > select FS_IOMAP > select DAX > diff --git a/fs/dax.c b/fs/dax.c > index b03f547b36e7..e0a3958fc5f2 100644 > --- a/fs/dax.c > +++ b/fs/dax.c > @@ -22,6 +22,7 @@ > #include <linux/genhd.h> > #include <linux/highmem.h> > #include <linux/memcontrol.h> > +#include <linux/file.h> > #include <linux/mm.h> > #include <linux/mutex.h> > #include <linux/pagevec.h> > @@ -1481,3 +1482,190 @@ int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size, > } > } > EXPORT_SYMBOL_GPL(dax_iomap_fault); > + > +enum dax_lease_flags { > + DAX_LEASE_PAGES, > + DAX_LEASE_BREAK, > +}; > + > +struct dax_lease { > + struct page **dl_pages; > + unsigned long dl_nr_pages; > + unsigned long dl_state; > + struct file *dl_file; > + atomic_t dl_count; > + /* > + * Once the lease is taken and the pages have references we > + * start the reap_work to poll for lease release while acquiring > + * fs locks that synchronize with truncate. So, either reap_work > + * cleans up the dax_lease instances or truncate itself. > + * > + * The break_work sleepily polls for DMA completion and then > + * unlocks/removes the lease. > + */ > + struct delayed_work dl_reap_work; > + struct delayed_work dl_break_work; > +}; > + > +static void put_dax_lease(struct dax_lease *dl) > +{ > + if (atomic_dec_and_test(&dl->dl_count)) { > + fput(dl->dl_file); > + kfree(dl); > + } > +} Any reason not to use the new refcount_t type for dl_count? Seems like a good place for it. > + > +static void dax_lease_unlock_one(struct work_struct *work) > +{ > + struct dax_lease *dl = container_of(work, typeof(*dl), > + dl_break_work.work); > + unsigned long i; > + > + /* wait for the gup path to finish recording pages in the lease */ > + if (!test_bit(DAX_LEASE_PAGES, &dl->dl_state)) { > + schedule_delayed_work(&dl->dl_break_work, HZ); > + return; > + } > + > + /* barrier pairs with dax_lease_set_pages() */ > + smp_mb__after_atomic(); > + > + /* > + * If we see all pages idle at least once we can remove the > + * lease. If we happen to race with someone else taking a > + * reference on a page they will have their own lease to protect > + * against truncate. > + */ > + for (i = 0; i < dl->dl_nr_pages; i++) > + if (page_ref_count(dl->dl_pages[i]) > 1) { > + schedule_delayed_work(&dl->dl_break_work, HZ); > + return; > + } > + vfs_setlease(dl->dl_file, F_UNLCK, NULL, (void **) &dl); > + put_dax_lease(dl); > +} > + > +static void dax_lease_reap_all(struct work_struct *work) > +{ > + struct dax_lease *dl = container_of(work, typeof(*dl), > + dl_reap_work.work); > + struct file *file = dl->dl_file; > + struct inode *inode = file_inode(file); > + struct address_space *mapping = inode->i_mapping; > + > + if (mapping->a_ops->dax_flush_dma) { > + mapping->a_ops->dax_flush_dma(inode); > + } else { > + /* FIXME: dax-filesystem needs to add dax-dma support */ > + break_allocated(inode, true); > + } > + put_dax_lease(dl); > +} > + > +static bool dax_lease_lm_break(struct file_lock *fl) > +{ > + struct dax_lease *dl = fl->fl_owner; > + > + if (!test_and_set_bit(DAX_LEASE_BREAK, &dl->dl_state)) { > + atomic_inc(&dl->dl_count); > + schedule_delayed_work(&dl->dl_break_work, HZ); > + } > + I haven't gone over this completely, but what prevents you from doing a 0->1 transition on the dl_count here, and possibly doing a use-after free? Ahh ok...I guess we know that we hold a reference since this is on the flc_lease list? Fair enough. Still, might be worth a comment there as to why that's safe. > + /* Tell the core lease code to wait for delayed work completion */ > + fl->fl_break_time = 0; > + > + return false; > +} > + > +static int dax_lease_lm_change(struct file_lock *fl, int arg, > + struct list_head *dispose) > +{ > + struct dax_lease *dl; > + int rc; > + > + WARN_ON(!(arg & F_UNLCK)); > + dl = fl->fl_owner; > + rc = lease_modify(fl, arg, dispose); > + put_dax_lease(dl); > + return rc; > +} > + > +static const struct lock_manager_operations dax_lease_lm_ops = { > + .lm_break = dax_lease_lm_break, > + .lm_change = dax_lease_lm_change, > +}; > + > +struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, > + long nr_pages) > +{ > + struct file *file = vma->vm_file; > + struct inode *inode = file_inode(file); > + struct dax_lease *dl; > + struct file_lock *fl; > + int rc = -ENOMEM; > + > + if (!vma_is_dax(vma)) > + return NULL; > + > + /* device-dax can not be truncated */ > + if (!S_ISREG(inode->i_mode)) > + return NULL; > + > + dl = kzalloc(sizeof(*dl) + sizeof(struct page *) * nr_pages, GFP_KERNEL); > + if (!dl) > + return ERR_PTR(-ENOMEM); > + > + fl = locks_alloc_lock(); > + if (!fl) > + goto err_lock_alloc; > + > + dl->dl_pages = (struct page **)(dl + 1); > + INIT_DELAYED_WORK(&dl->dl_break_work, dax_lease_unlock_one); > + INIT_DELAYED_WORK(&dl->dl_reap_work, dax_lease_reap_all); > + dl->dl_file = get_file(file); > + /* need dl alive until dax_lease_set_pages() and final put */ > + atomic_set(&dl->dl_count, 2); > + > + locks_init_lock(fl); > + fl->fl_lmops = &dax_lease_lm_ops; > + fl->fl_flags = FL_ALLOCATED; > + fl->fl_type = F_RDLCK; > + fl->fl_end = OFFSET_MAX; > + fl->fl_owner = dl; > + fl->fl_pid = current->tgid; > + fl->fl_file = file; > + > + rc = vfs_setlease(fl->fl_file, fl->fl_type, &fl, (void **) &dl); > + if (rc) > + goto err_setlease; > + return dl; > +err_setlease: > + locks_free_lock(fl); > +err_lock_alloc: > + kfree(dl); > + return ERR_PTR(rc); > +} > + > +void dax_lease_set_pages(struct dax_lease *dl, struct page **pages, > + long nr_pages) > +{ > + if (IS_ERR_OR_NULL(dl)) > + return; > + > + if (nr_pages <= 0) { > + dl->dl_nr_pages = 0; > + smp_mb__before_atomic(); > + set_bit(DAX_LEASE_PAGES, &dl->dl_state); > + vfs_setlease(dl->dl_file, F_UNLCK, NULL, (void **) &dl); > + flush_delayed_work(&dl->dl_break_work); > + put_dax_lease(dl); > + return; > + } > + > + dl->dl_nr_pages = nr_pages; > + memcpy(dl->dl_pages, pages, sizeof(struct page *) * nr_pages); > + smp_mb__before_atomic(); > + set_bit(DAX_LEASE_PAGES, &dl->dl_state); > + queue_delayed_work(system_long_wq, &dl->dl_reap_work, HZ); > +} > +EXPORT_SYMBOL_GPL(dax_lease_set_pages); > diff --git a/fs/locks.c b/fs/locks.c > index 1bd71c4d663a..0a7841590b35 100644 > --- a/fs/locks.c > +++ b/fs/locks.c > @@ -135,7 +135,7 @@ > > #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) > #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) > -#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) > +#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT|FL_ALLOCATED)) > #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) > #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) > > @@ -1414,7 +1414,9 @@ static void time_out_leases(struct inode *inode, struct list_head *dispose) > > static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) > { > - if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) > + /* FL_LAYOUT and FL_ALLOCATED only conflict with each other */ > + if (!!(breaker->fl_flags & (FL_LAYOUT|FL_ALLOCATED)) > + != !!(lease->fl_flags & (FL_LAYOUT|FL_ALLOCATED))) > return false; > if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) > return false; > @@ -1653,7 +1655,7 @@ check_conflicting_open(const struct dentry *dentry, const long arg, int flags) > int ret = 0; > struct inode *inode = dentry->d_inode; > > - if (flags & FL_LAYOUT) > + if (flags & (FL_LAYOUT|FL_ALLOCATED)) > return 0; > > if ((arg == F_RDLCK) && > @@ -1733,6 +1735,15 @@ generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **pr > */ > if (arg == F_WRLCK) > goto out; > + > + /* > + * Taking out a new FL_ALLOCATED lease while a previous > + * one is being locked is expected since each instance > + * may be responsible for a distinct range of pages. > + */ > + if (fl->fl_flags & FL_ALLOCATED) > + continue; > + > /* > * Modifying our existing lease is OK, but no getting a > * new lease if someone else is opening for write: > diff --git a/include/linux/dax.h b/include/linux/dax.h > index 122197124b9d..3ff61dc6241e 100644 > --- a/include/linux/dax.h > +++ b/include/linux/dax.h > @@ -100,10 +100,15 @@ int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index); > int dax_invalidate_mapping_entry_sync(struct address_space *mapping, > pgoff_t index); > > +struct dax_lease; > #ifdef CONFIG_FS_DAX > int __dax_zero_page_range(struct block_device *bdev, > struct dax_device *dax_dev, sector_t sector, > unsigned int offset, unsigned int length); > +struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, > + long nr_pages); > +void dax_lease_set_pages(struct dax_lease *dl, struct page **pages, > + long nr_pages); > #else > static inline int __dax_zero_page_range(struct block_device *bdev, > struct dax_device *dax_dev, sector_t sector, > @@ -111,8 +116,26 @@ static inline int __dax_zero_page_range(struct block_device *bdev, > { > return -ENXIO; > } > +static inline struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, > + long nr_pages) > +{ > + return NULL; > +} > + > +static inline void dax_lease_set_pages(struct dax_lease *dl, > + struct page **pages, long nr_pages) > +{ > +} > #endif > > +static inline struct dax_lease *dax_truncate_lease(struct vm_area_struct *vma, > + long nr_pages) > +{ > + if (!vma_is_dax(vma)) > + return NULL; > + return __dax_truncate_lease(vma, nr_pages); > +} > + > static inline bool dax_mapping(struct address_space *mapping) > { > return mapping->host && IS_DAX(mapping->host); > diff --git a/include/linux/fs.h b/include/linux/fs.h > index eace2c5396a7..a3ed74833919 100644 > --- a/include/linux/fs.h > +++ b/include/linux/fs.h > @@ -371,6 +371,9 @@ struct address_space_operations { > int (*swap_activate)(struct swap_info_struct *sis, struct file *file, > sector_t *span); > void (*swap_deactivate)(struct file *file); > + > + /* dax dma support */ > + void (*dax_flush_dma)(struct inode *inode); > }; > > extern const struct address_space_operations empty_aops; > @@ -927,6 +930,7 @@ static inline struct file *get_file(struct file *f) > #define FL_UNLOCK_PENDING 512 /* Lease is being broken */ > #define FL_OFDLCK 1024 /* lock is "owned" by struct file */ > #define FL_LAYOUT 2048 /* outstanding pNFS layout */ > +#define FL_ALLOCATED 4096 /* pin allocated dax blocks against dma */ > > #define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE) > > @@ -2324,17 +2328,27 @@ static inline int break_deleg_wait(struct inode **delegated_inode) > return ret; > } > > -static inline int break_layout(struct inode *inode, bool wait) > +static inline int __break_layout(struct inode *inode, bool wait, > + unsigned int type) > { > struct file_lock_context *ctx = smp_load_acquire(&inode->i_flctx); > > if (ctx && !list_empty_careful(&ctx->flc_lease)) > return __break_lease(inode, > wait ? O_WRONLY : O_WRONLY | O_NONBLOCK, > - FL_LAYOUT); > + type); > return 0; > } > > +static inline int break_layout(struct inode *inode, bool wait) > +{ > + return __break_layout(inode, wait, FL_LAYOUT); > +} > + > +static inline int break_allocated(struct inode *inode, bool wait) > +{ > + return __break_layout(inode, wait, FL_LAYOUT|FL_ALLOCATED); > +} > #else /* !CONFIG_FILE_LOCKING */ > static inline int break_lease(struct inode *inode, unsigned int mode) > { > @@ -2362,6 +2376,10 @@ static inline int break_layout(struct inode *inode, bool wait) > return 0; > } > > +static inline int break_allocated(struct inode *inode, bool wait) > +{ > + return 0; > +} > #endif /* CONFIG_FILE_LOCKING */ > > /* fs/open.c */ > diff --git a/mm/gup.c b/mm/gup.c > index 308be897d22a..6a7cf371e656 100644 > --- a/mm/gup.c > +++ b/mm/gup.c > @@ -9,6 +9,7 @@ > #include <linux/rmap.h> > #include <linux/swap.h> > #include <linux/swapops.h> > +#include <linux/dax.h> > > #include <linux/sched/signal.h> > #include <linux/rwsem.h> > @@ -640,9 +641,11 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, > unsigned int gup_flags, struct page **pages, > struct vm_area_struct **vmas, int *nonblocking) > { > - long i = 0; > + long i = 0, result = 0; > + int dax_lease_once = 0; > unsigned int page_mask; > struct vm_area_struct *vma = NULL; > + struct dax_lease *dax_lease = NULL; > > if (!nr_pages) > return 0; > @@ -693,6 +696,14 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, > if (unlikely(fatal_signal_pending(current))) > return i ? i : -ERESTARTSYS; > cond_resched(); > + if (pages && !dax_lease_once) { > + dax_lease_once = 1; > + dax_lease = dax_truncate_lease(vma, nr_pages); > + if (IS_ERR(dax_lease)) { > + result = PTR_ERR(dax_lease); > + goto out; > + } > + } > page = follow_page_mask(vma, start, foll_flags, &page_mask); > if (!page) { > int ret; > @@ -704,9 +715,11 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, > case -EFAULT: > case -ENOMEM: > case -EHWPOISON: > - return i ? i : ret; > + result = i ? i : ret; > + goto out; > case -EBUSY: > - return i; > + result = i; > + goto out; > case -ENOENT: > goto next_page; > } > @@ -718,7 +731,8 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, > */ > goto next_page; > } else if (IS_ERR(page)) { > - return i ? i : PTR_ERR(page); > + result = i ? i : PTR_ERR(page); > + goto out; > } > if (pages) { > pages[i] = page; > @@ -738,7 +752,10 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, > start += page_increm * PAGE_SIZE; > nr_pages -= page_increm; > } while (nr_pages); > - return i; > + result = i; > +out: > + dax_lease_set_pages(dax_lease, pages, result); > + return result; > } > > static bool vma_permits_fault(struct vm_area_struct *vma, >
On Fri, Oct 20, 2017 at 6:05 AM, Jeff Layton <jlayton@kernel.org> wrote: > On Thu, 2017-10-19 at 19:40 -0700, Dan Williams wrote: >> get_user_pages() pins file backed memory pages for access by dma >> devices. However, it only pins the memory pages not the page-to-file >> offset association. If a file is truncated the pages are mapped out of >> the file and dma may continue indefinitely into a page that is owned by >> a device driver. This breaks coherency of the file vs dma, but the >> assumption is that if userspace wants the file-space truncated it does >> not matter what data is inbound from the device, it is not relevant >> anymore. >> >> The assumptions of the truncate-page-cache model are broken by DAX where >> the target DMA page *is* the filesystem block. Leaving the page pinned >> for DMA, but truncating the file block out of the file, means that the >> filesytem is free to reallocate a block under active DMA to another >> file! >> >> Here are some possible options for fixing this situation ('truncate' and >> 'fallocate(punch hole)' are synonymous below): >> >> 1/ Fail truncate while any file blocks might be under dma >> >> 2/ Block (sleep-wait) truncate while any file blocks might be under >> dma >> >> 3/ Remap file blocks to a "lost+found"-like file-inode where >> dma can continue and we might see what inbound data from DMA was >> mapped out of the original file. Blocks in this file could be >> freed back to the filesystem when dma eventually ends. >> >> 4/ Disable dax until option 3 or another long term solution has been >> implemented. However, filesystem-dax is still marked experimental >> for concerns like this. >> >> Option 1 will throw failures where userspace has never expected them >> before, option 2 might hang the truncating process indefinitely, and >> option 3 requires per filesystem enabling to remap blocks from one inode >> to another. Option 2 is implemented in this patch for the DAX path with >> the expectation that non-transient users of get_user_pages() (RDMA) are >> disallowed from setting up dax mappings and that the potential delay >> introduced to the truncate path is acceptable compared to the response >> time of the page cache case. This can only be seen as a stop-gap until >> we can solve the problem of safely sequestering unallocated filesystem >> blocks under active dma. >> > > FWIW, I like #3 a lot more than #2 here. I get that it's quite a bit > more work though, so no objection to this as a stop-gap fix. I agree, but it needs quite a bit more thought and restructuring of the truncate path. I also wonder how we reclaim those stranded filesystem blocks, but a first approximation is wait for the administrator to delete them or auto-delete them at the next mount. XFS seems well prepared to reflink-swap these DMA blocks around, but I'm not sure about EXT4. > > >> The solution introduces a new FL_ALLOCATED lease to pin the allocated >> blocks in a dax file while dma might be accessing them. It behaves >> identically to an FL_LAYOUT lease save for the fact that it is >> immediately sheduled to be reaped, and that the only path that waits for >> its removal is the truncate path. We can not reuse FL_LAYOUT directly >> since that would deadlock in the case where userspace did a direct-I/O >> operation with a target buffer backed by an mmap range of the same file. >> >> Credit / inspiration for option 3 goes to Dave Hansen, who proposed >> something similar as an alternative way to solve the problem that >> MAP_DIRECT was trying to solve. >> >> Cc: Jan Kara <jack@suse.cz> >> Cc: Jeff Moyer <jmoyer@redhat.com> >> Cc: Dave Chinner <david@fromorbit.com> >> Cc: Matthew Wilcox <mawilcox@microsoft.com> >> Cc: Alexander Viro <viro@zeniv.linux.org.uk> >> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> >> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> >> Cc: Jeff Layton <jlayton@poochiereds.net> >> Cc: "J. Bruce Fields" <bfields@fieldses.org> >> Cc: Dave Hansen <dave.hansen@linux.intel.com> >> Reported-by: Christoph Hellwig <hch@lst.de> >> Signed-off-by: Dan Williams <dan.j.williams@intel.com> >> --- >> fs/Kconfig | 1 >> fs/dax.c | 188 +++++++++++++++++++++++++++++++++++++++++++++++++++ >> fs/locks.c | 17 ++++- >> include/linux/dax.h | 23 ++++++ >> include/linux/fs.h | 22 +++++- >> mm/gup.c | 27 ++++++- >> 6 files changed, 268 insertions(+), 10 deletions(-) >> >> diff --git a/fs/Kconfig b/fs/Kconfig >> index 7aee6d699fd6..a7b31a96a753 100644 >> --- a/fs/Kconfig >> +++ b/fs/Kconfig >> @@ -37,6 +37,7 @@ source "fs/f2fs/Kconfig" >> config FS_DAX >> bool "Direct Access (DAX) support" >> depends on MMU >> + depends on FILE_LOCKING >> depends on !(ARM || MIPS || SPARC) >> select FS_IOMAP >> select DAX >> diff --git a/fs/dax.c b/fs/dax.c >> index b03f547b36e7..e0a3958fc5f2 100644 >> --- a/fs/dax.c >> +++ b/fs/dax.c >> @@ -22,6 +22,7 @@ >> #include <linux/genhd.h> >> #include <linux/highmem.h> >> #include <linux/memcontrol.h> >> +#include <linux/file.h> >> #include <linux/mm.h> >> #include <linux/mutex.h> >> #include <linux/pagevec.h> >> @@ -1481,3 +1482,190 @@ int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size, >> } >> } >> EXPORT_SYMBOL_GPL(dax_iomap_fault); >> + >> +enum dax_lease_flags { >> + DAX_LEASE_PAGES, >> + DAX_LEASE_BREAK, >> +}; >> + >> +struct dax_lease { >> + struct page **dl_pages; >> + unsigned long dl_nr_pages; >> + unsigned long dl_state; >> + struct file *dl_file; >> + atomic_t dl_count; >> + /* >> + * Once the lease is taken and the pages have references we >> + * start the reap_work to poll for lease release while acquiring >> + * fs locks that synchronize with truncate. So, either reap_work >> + * cleans up the dax_lease instances or truncate itself. >> + * >> + * The break_work sleepily polls for DMA completion and then >> + * unlocks/removes the lease. >> + */ >> + struct delayed_work dl_reap_work; >> + struct delayed_work dl_break_work; >> +}; >> + >> +static void put_dax_lease(struct dax_lease *dl) >> +{ >> + if (atomic_dec_and_test(&dl->dl_count)) { >> + fput(dl->dl_file); >> + kfree(dl); >> + } >> +} > > Any reason not to use the new refcount_t type for dl_count? Seems like a > good place for it. I'll take a look. >> + >> +static void dax_lease_unlock_one(struct work_struct *work) >> +{ >> + struct dax_lease *dl = container_of(work, typeof(*dl), >> + dl_break_work.work); >> + unsigned long i; >> + >> + /* wait for the gup path to finish recording pages in the lease */ >> + if (!test_bit(DAX_LEASE_PAGES, &dl->dl_state)) { >> + schedule_delayed_work(&dl->dl_break_work, HZ); >> + return; >> + } >> + >> + /* barrier pairs with dax_lease_set_pages() */ >> + smp_mb__after_atomic(); >> + >> + /* >> + * If we see all pages idle at least once we can remove the >> + * lease. If we happen to race with someone else taking a >> + * reference on a page they will have their own lease to protect >> + * against truncate. >> + */ >> + for (i = 0; i < dl->dl_nr_pages; i++) >> + if (page_ref_count(dl->dl_pages[i]) > 1) { >> + schedule_delayed_work(&dl->dl_break_work, HZ); >> + return; >> + } >> + vfs_setlease(dl->dl_file, F_UNLCK, NULL, (void **) &dl); >> + put_dax_lease(dl); >> +} >> + >> +static void dax_lease_reap_all(struct work_struct *work) >> +{ >> + struct dax_lease *dl = container_of(work, typeof(*dl), >> + dl_reap_work.work); >> + struct file *file = dl->dl_file; >> + struct inode *inode = file_inode(file); >> + struct address_space *mapping = inode->i_mapping; >> + >> + if (mapping->a_ops->dax_flush_dma) { >> + mapping->a_ops->dax_flush_dma(inode); >> + } else { >> + /* FIXME: dax-filesystem needs to add dax-dma support */ >> + break_allocated(inode, true); >> + } >> + put_dax_lease(dl); >> +} >> + >> +static bool dax_lease_lm_break(struct file_lock *fl) >> +{ >> + struct dax_lease *dl = fl->fl_owner; >> + >> + if (!test_and_set_bit(DAX_LEASE_BREAK, &dl->dl_state)) { >> + atomic_inc(&dl->dl_count); >> + schedule_delayed_work(&dl->dl_break_work, HZ); >> + } >> + > > I haven't gone over this completely, but what prevents you from doing a > 0->1 transition on the dl_count here, and possibly doing a use-after > free? > > Ahh ok...I guess we know that we hold a reference since this is on the > flc_lease list? Fair enough. Still, might be worth a comment there as to > why that's safe. Right, we hold a reference count at the beginning of time that is only dropped when the lease is unlocked. If the break happens before unlock we take this reference while the break_work is running. I'll add this as a comment.
On Fri, Oct 20, 2017 at 08:42:00AM -0700, Dan Williams wrote: > I agree, but it needs quite a bit more thought and restructuring of > the truncate path. I also wonder how we reclaim those stranded > filesystem blocks, but a first approximation is wait for the > administrator to delete them or auto-delete them at the next mount. > XFS seems well prepared to reflink-swap these DMA blocks around, but > I'm not sure about EXT4. reflink still is an optional and experimental feature in XFS. That being said we should not need to swap block pointers around on disk. We just need to prevent the block allocator from reusing the blocks for new allocations, and we have code for that, both for transactions that haven't been committed to disk yet, and for deleted blocks undergoing discard operations. But as mentioned in my second mail from this morning I'm not even sure we need that. For short-term elevated page counts like normal get_user_pages users I think we can just wait for the page count to reach zero, while for abuses of get_user_pages for long term pinning memory (not sure if anyone but rdma is doing that) we'll need something like FL_LAYOUT leases to release the mapping.
On Fri, Oct 20, 2017 at 9:32 AM, Christoph Hellwig <hch@lst.de> wrote: > On Fri, Oct 20, 2017 at 08:42:00AM -0700, Dan Williams wrote: >> I agree, but it needs quite a bit more thought and restructuring of >> the truncate path. I also wonder how we reclaim those stranded >> filesystem blocks, but a first approximation is wait for the >> administrator to delete them or auto-delete them at the next mount. >> XFS seems well prepared to reflink-swap these DMA blocks around, but >> I'm not sure about EXT4. > > reflink still is an optional and experimental feature in XFS. That > being said we should not need to swap block pointers around on disk. > We just need to prevent the block allocator from reusing the blocks > for new allocations, and we have code for that, both for transactions > that haven't been committed to disk yet, and for deleted blocks > undergoing discard operations. > > But as mentioned in my second mail from this morning I'm not even > sure we need that. For short-term elevated page counts like normal > get_user_pages users I think we can just wait for the page count > to reach zero, while for abuses of get_user_pages for long term > pinning memory (not sure if anyone but rdma is doing that) we'll need > something like FL_LAYOUT leases to release the mapping. I'll take a look at hooking this up through a page-idle callback. Can I get some breadcrumbs to grep for from XFS folks on how to set/clear the busy state of extents?
On Fri, Oct 20, 2017 at 10:27:22AM -0700, Dan Williams wrote: > On Fri, Oct 20, 2017 at 9:32 AM, Christoph Hellwig <hch@lst.de> wrote: > > On Fri, Oct 20, 2017 at 08:42:00AM -0700, Dan Williams wrote: > >> I agree, but it needs quite a bit more thought and restructuring of > >> the truncate path. I also wonder how we reclaim those stranded > >> filesystem blocks, but a first approximation is wait for the > >> administrator to delete them or auto-delete them at the next mount. > >> XFS seems well prepared to reflink-swap these DMA blocks around, but > >> I'm not sure about EXT4. > > > > reflink still is an optional and experimental feature in XFS. That > > being said we should not need to swap block pointers around on disk. > > We just need to prevent the block allocator from reusing the blocks > > for new allocations, and we have code for that, both for transactions > > that haven't been committed to disk yet, and for deleted blocks > > undergoing discard operations. > > > > But as mentioned in my second mail from this morning I'm not even > > sure we need that. For short-term elevated page counts like normal > > get_user_pages users I think we can just wait for the page count > > to reach zero, while for abuses of get_user_pages for long term > > pinning memory (not sure if anyone but rdma is doing that) we'll need > > something like FL_LAYOUT leases to release the mapping. > > I'll take a look at hooking this up through a page-idle callback. Can > I get some breadcrumbs to grep for from XFS folks on how to set/clear > the busy state of extents? See fs/xfs/xfs_extent_busy.c. Brian
On Fri, Oct 20, 2017 at 10:27:22AM -0700, Dan Williams wrote: > I'll take a look at hooking this up through a page-idle callback. Can > I get some breadcrumbs to grep for from XFS folks on how to set/clear > the busy state of extents? As Brian pointed out it's the xfs_extent_busy.c file (and I pointed out the same in a reply to the previous series). Be careful because you'll need a refcount or flags now that there are different busy reasons. I still think we'd be better off just blocking on an elevated page count directly in truncate as that will avoid all the busy list manipulations.
diff --git a/fs/Kconfig b/fs/Kconfig index 7aee6d699fd6..a7b31a96a753 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -37,6 +37,7 @@ source "fs/f2fs/Kconfig" config FS_DAX bool "Direct Access (DAX) support" depends on MMU + depends on FILE_LOCKING depends on !(ARM || MIPS || SPARC) select FS_IOMAP select DAX diff --git a/fs/dax.c b/fs/dax.c index b03f547b36e7..e0a3958fc5f2 100644 --- a/fs/dax.c +++ b/fs/dax.c @@ -22,6 +22,7 @@ #include <linux/genhd.h> #include <linux/highmem.h> #include <linux/memcontrol.h> +#include <linux/file.h> #include <linux/mm.h> #include <linux/mutex.h> #include <linux/pagevec.h> @@ -1481,3 +1482,190 @@ int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size, } } EXPORT_SYMBOL_GPL(dax_iomap_fault); + +enum dax_lease_flags { + DAX_LEASE_PAGES, + DAX_LEASE_BREAK, +}; + +struct dax_lease { + struct page **dl_pages; + unsigned long dl_nr_pages; + unsigned long dl_state; + struct file *dl_file; + atomic_t dl_count; + /* + * Once the lease is taken and the pages have references we + * start the reap_work to poll for lease release while acquiring + * fs locks that synchronize with truncate. So, either reap_work + * cleans up the dax_lease instances or truncate itself. + * + * The break_work sleepily polls for DMA completion and then + * unlocks/removes the lease. + */ + struct delayed_work dl_reap_work; + struct delayed_work dl_break_work; +}; + +static void put_dax_lease(struct dax_lease *dl) +{ + if (atomic_dec_and_test(&dl->dl_count)) { + fput(dl->dl_file); + kfree(dl); + } +} + +static void dax_lease_unlock_one(struct work_struct *work) +{ + struct dax_lease *dl = container_of(work, typeof(*dl), + dl_break_work.work); + unsigned long i; + + /* wait for the gup path to finish recording pages in the lease */ + if (!test_bit(DAX_LEASE_PAGES, &dl->dl_state)) { + schedule_delayed_work(&dl->dl_break_work, HZ); + return; + } + + /* barrier pairs with dax_lease_set_pages() */ + smp_mb__after_atomic(); + + /* + * If we see all pages idle at least once we can remove the + * lease. If we happen to race with someone else taking a + * reference on a page they will have their own lease to protect + * against truncate. + */ + for (i = 0; i < dl->dl_nr_pages; i++) + if (page_ref_count(dl->dl_pages[i]) > 1) { + schedule_delayed_work(&dl->dl_break_work, HZ); + return; + } + vfs_setlease(dl->dl_file, F_UNLCK, NULL, (void **) &dl); + put_dax_lease(dl); +} + +static void dax_lease_reap_all(struct work_struct *work) +{ + struct dax_lease *dl = container_of(work, typeof(*dl), + dl_reap_work.work); + struct file *file = dl->dl_file; + struct inode *inode = file_inode(file); + struct address_space *mapping = inode->i_mapping; + + if (mapping->a_ops->dax_flush_dma) { + mapping->a_ops->dax_flush_dma(inode); + } else { + /* FIXME: dax-filesystem needs to add dax-dma support */ + break_allocated(inode, true); + } + put_dax_lease(dl); +} + +static bool dax_lease_lm_break(struct file_lock *fl) +{ + struct dax_lease *dl = fl->fl_owner; + + if (!test_and_set_bit(DAX_LEASE_BREAK, &dl->dl_state)) { + atomic_inc(&dl->dl_count); + schedule_delayed_work(&dl->dl_break_work, HZ); + } + + /* Tell the core lease code to wait for delayed work completion */ + fl->fl_break_time = 0; + + return false; +} + +static int dax_lease_lm_change(struct file_lock *fl, int arg, + struct list_head *dispose) +{ + struct dax_lease *dl; + int rc; + + WARN_ON(!(arg & F_UNLCK)); + dl = fl->fl_owner; + rc = lease_modify(fl, arg, dispose); + put_dax_lease(dl); + return rc; +} + +static const struct lock_manager_operations dax_lease_lm_ops = { + .lm_break = dax_lease_lm_break, + .lm_change = dax_lease_lm_change, +}; + +struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, + long nr_pages) +{ + struct file *file = vma->vm_file; + struct inode *inode = file_inode(file); + struct dax_lease *dl; + struct file_lock *fl; + int rc = -ENOMEM; + + if (!vma_is_dax(vma)) + return NULL; + + /* device-dax can not be truncated */ + if (!S_ISREG(inode->i_mode)) + return NULL; + + dl = kzalloc(sizeof(*dl) + sizeof(struct page *) * nr_pages, GFP_KERNEL); + if (!dl) + return ERR_PTR(-ENOMEM); + + fl = locks_alloc_lock(); + if (!fl) + goto err_lock_alloc; + + dl->dl_pages = (struct page **)(dl + 1); + INIT_DELAYED_WORK(&dl->dl_break_work, dax_lease_unlock_one); + INIT_DELAYED_WORK(&dl->dl_reap_work, dax_lease_reap_all); + dl->dl_file = get_file(file); + /* need dl alive until dax_lease_set_pages() and final put */ + atomic_set(&dl->dl_count, 2); + + locks_init_lock(fl); + fl->fl_lmops = &dax_lease_lm_ops; + fl->fl_flags = FL_ALLOCATED; + fl->fl_type = F_RDLCK; + fl->fl_end = OFFSET_MAX; + fl->fl_owner = dl; + fl->fl_pid = current->tgid; + fl->fl_file = file; + + rc = vfs_setlease(fl->fl_file, fl->fl_type, &fl, (void **) &dl); + if (rc) + goto err_setlease; + return dl; +err_setlease: + locks_free_lock(fl); +err_lock_alloc: + kfree(dl); + return ERR_PTR(rc); +} + +void dax_lease_set_pages(struct dax_lease *dl, struct page **pages, + long nr_pages) +{ + if (IS_ERR_OR_NULL(dl)) + return; + + if (nr_pages <= 0) { + dl->dl_nr_pages = 0; + smp_mb__before_atomic(); + set_bit(DAX_LEASE_PAGES, &dl->dl_state); + vfs_setlease(dl->dl_file, F_UNLCK, NULL, (void **) &dl); + flush_delayed_work(&dl->dl_break_work); + put_dax_lease(dl); + return; + } + + dl->dl_nr_pages = nr_pages; + memcpy(dl->dl_pages, pages, sizeof(struct page *) * nr_pages); + smp_mb__before_atomic(); + set_bit(DAX_LEASE_PAGES, &dl->dl_state); + queue_delayed_work(system_long_wq, &dl->dl_reap_work, HZ); +} +EXPORT_SYMBOL_GPL(dax_lease_set_pages); diff --git a/fs/locks.c b/fs/locks.c index 1bd71c4d663a..0a7841590b35 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -135,7 +135,7 @@ #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) -#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) +#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT|FL_ALLOCATED)) #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) @@ -1414,7 +1414,9 @@ static void time_out_leases(struct inode *inode, struct list_head *dispose) static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) { - if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) + /* FL_LAYOUT and FL_ALLOCATED only conflict with each other */ + if (!!(breaker->fl_flags & (FL_LAYOUT|FL_ALLOCATED)) + != !!(lease->fl_flags & (FL_LAYOUT|FL_ALLOCATED))) return false; if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) return false; @@ -1653,7 +1655,7 @@ check_conflicting_open(const struct dentry *dentry, const long arg, int flags) int ret = 0; struct inode *inode = dentry->d_inode; - if (flags & FL_LAYOUT) + if (flags & (FL_LAYOUT|FL_ALLOCATED)) return 0; if ((arg == F_RDLCK) && @@ -1733,6 +1735,15 @@ generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **pr */ if (arg == F_WRLCK) goto out; + + /* + * Taking out a new FL_ALLOCATED lease while a previous + * one is being locked is expected since each instance + * may be responsible for a distinct range of pages. + */ + if (fl->fl_flags & FL_ALLOCATED) + continue; + /* * Modifying our existing lease is OK, but no getting a * new lease if someone else is opening for write: diff --git a/include/linux/dax.h b/include/linux/dax.h index 122197124b9d..3ff61dc6241e 100644 --- a/include/linux/dax.h +++ b/include/linux/dax.h @@ -100,10 +100,15 @@ int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index); int dax_invalidate_mapping_entry_sync(struct address_space *mapping, pgoff_t index); +struct dax_lease; #ifdef CONFIG_FS_DAX int __dax_zero_page_range(struct block_device *bdev, struct dax_device *dax_dev, sector_t sector, unsigned int offset, unsigned int length); +struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, + long nr_pages); +void dax_lease_set_pages(struct dax_lease *dl, struct page **pages, + long nr_pages); #else static inline int __dax_zero_page_range(struct block_device *bdev, struct dax_device *dax_dev, sector_t sector, @@ -111,8 +116,26 @@ static inline int __dax_zero_page_range(struct block_device *bdev, { return -ENXIO; } +static inline struct dax_lease *__dax_truncate_lease(struct vm_area_struct *vma, + long nr_pages) +{ + return NULL; +} + +static inline void dax_lease_set_pages(struct dax_lease *dl, + struct page **pages, long nr_pages) +{ +} #endif +static inline struct dax_lease *dax_truncate_lease(struct vm_area_struct *vma, + long nr_pages) +{ + if (!vma_is_dax(vma)) + return NULL; + return __dax_truncate_lease(vma, nr_pages); +} + static inline bool dax_mapping(struct address_space *mapping) { return mapping->host && IS_DAX(mapping->host); diff --git a/include/linux/fs.h b/include/linux/fs.h index eace2c5396a7..a3ed74833919 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -371,6 +371,9 @@ struct address_space_operations { int (*swap_activate)(struct swap_info_struct *sis, struct file *file, sector_t *span); void (*swap_deactivate)(struct file *file); + + /* dax dma support */ + void (*dax_flush_dma)(struct inode *inode); }; extern const struct address_space_operations empty_aops; @@ -927,6 +930,7 @@ static inline struct file *get_file(struct file *f) #define FL_UNLOCK_PENDING 512 /* Lease is being broken */ #define FL_OFDLCK 1024 /* lock is "owned" by struct file */ #define FL_LAYOUT 2048 /* outstanding pNFS layout */ +#define FL_ALLOCATED 4096 /* pin allocated dax blocks against dma */ #define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE) @@ -2324,17 +2328,27 @@ static inline int break_deleg_wait(struct inode **delegated_inode) return ret; } -static inline int break_layout(struct inode *inode, bool wait) +static inline int __break_layout(struct inode *inode, bool wait, + unsigned int type) { struct file_lock_context *ctx = smp_load_acquire(&inode->i_flctx); if (ctx && !list_empty_careful(&ctx->flc_lease)) return __break_lease(inode, wait ? O_WRONLY : O_WRONLY | O_NONBLOCK, - FL_LAYOUT); + type); return 0; } +static inline int break_layout(struct inode *inode, bool wait) +{ + return __break_layout(inode, wait, FL_LAYOUT); +} + +static inline int break_allocated(struct inode *inode, bool wait) +{ + return __break_layout(inode, wait, FL_LAYOUT|FL_ALLOCATED); +} #else /* !CONFIG_FILE_LOCKING */ static inline int break_lease(struct inode *inode, unsigned int mode) { @@ -2362,6 +2376,10 @@ static inline int break_layout(struct inode *inode, bool wait) return 0; } +static inline int break_allocated(struct inode *inode, bool wait) +{ + return 0; +} #endif /* CONFIG_FILE_LOCKING */ /* fs/open.c */ diff --git a/mm/gup.c b/mm/gup.c index 308be897d22a..6a7cf371e656 100644 --- a/mm/gup.c +++ b/mm/gup.c @@ -9,6 +9,7 @@ #include <linux/rmap.h> #include <linux/swap.h> #include <linux/swapops.h> +#include <linux/dax.h> #include <linux/sched/signal.h> #include <linux/rwsem.h> @@ -640,9 +641,11 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned int gup_flags, struct page **pages, struct vm_area_struct **vmas, int *nonblocking) { - long i = 0; + long i = 0, result = 0; + int dax_lease_once = 0; unsigned int page_mask; struct vm_area_struct *vma = NULL; + struct dax_lease *dax_lease = NULL; if (!nr_pages) return 0; @@ -693,6 +696,14 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, if (unlikely(fatal_signal_pending(current))) return i ? i : -ERESTARTSYS; cond_resched(); + if (pages && !dax_lease_once) { + dax_lease_once = 1; + dax_lease = dax_truncate_lease(vma, nr_pages); + if (IS_ERR(dax_lease)) { + result = PTR_ERR(dax_lease); + goto out; + } + } page = follow_page_mask(vma, start, foll_flags, &page_mask); if (!page) { int ret; @@ -704,9 +715,11 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, case -EFAULT: case -ENOMEM: case -EHWPOISON: - return i ? i : ret; + result = i ? i : ret; + goto out; case -EBUSY: - return i; + result = i; + goto out; case -ENOENT: goto next_page; } @@ -718,7 +731,8 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, */ goto next_page; } else if (IS_ERR(page)) { - return i ? i : PTR_ERR(page); + result = i ? i : PTR_ERR(page); + goto out; } if (pages) { pages[i] = page; @@ -738,7 +752,10 @@ static long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, start += page_increm * PAGE_SIZE; nr_pages -= page_increm; } while (nr_pages); - return i; + result = i; +out: + dax_lease_set_pages(dax_lease, pages, result); + return result; } static bool vma_permits_fault(struct vm_area_struct *vma,
get_user_pages() pins file backed memory pages for access by dma devices. However, it only pins the memory pages not the page-to-file offset association. If a file is truncated the pages are mapped out of the file and dma may continue indefinitely into a page that is owned by a device driver. This breaks coherency of the file vs dma, but the assumption is that if userspace wants the file-space truncated it does not matter what data is inbound from the device, it is not relevant anymore. The assumptions of the truncate-page-cache model are broken by DAX where the target DMA page *is* the filesystem block. Leaving the page pinned for DMA, but truncating the file block out of the file, means that the filesytem is free to reallocate a block under active DMA to another file! Here are some possible options for fixing this situation ('truncate' and 'fallocate(punch hole)' are synonymous below): 1/ Fail truncate while any file blocks might be under dma 2/ Block (sleep-wait) truncate while any file blocks might be under dma 3/ Remap file blocks to a "lost+found"-like file-inode where dma can continue and we might see what inbound data from DMA was mapped out of the original file. Blocks in this file could be freed back to the filesystem when dma eventually ends. 4/ Disable dax until option 3 or another long term solution has been implemented. However, filesystem-dax is still marked experimental for concerns like this. Option 1 will throw failures where userspace has never expected them before, option 2 might hang the truncating process indefinitely, and option 3 requires per filesystem enabling to remap blocks from one inode to another. Option 2 is implemented in this patch for the DAX path with the expectation that non-transient users of get_user_pages() (RDMA) are disallowed from setting up dax mappings and that the potential delay introduced to the truncate path is acceptable compared to the response time of the page cache case. This can only be seen as a stop-gap until we can solve the problem of safely sequestering unallocated filesystem blocks under active dma. The solution introduces a new FL_ALLOCATED lease to pin the allocated blocks in a dax file while dma might be accessing them. It behaves identically to an FL_LAYOUT lease save for the fact that it is immediately sheduled to be reaped, and that the only path that waits for its removal is the truncate path. We can not reuse FL_LAYOUT directly since that would deadlock in the case where userspace did a direct-I/O operation with a target buffer backed by an mmap range of the same file. Credit / inspiration for option 3 goes to Dave Hansen, who proposed something similar as an alternative way to solve the problem that MAP_DIRECT was trying to solve. Cc: Jan Kara <jack@suse.cz> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Dave Hansen <dave.hansen@linux.intel.com> Reported-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- fs/Kconfig | 1 fs/dax.c | 188 +++++++++++++++++++++++++++++++++++++++++++++++++++ fs/locks.c | 17 ++++- include/linux/dax.h | 23 ++++++ include/linux/fs.h | 22 +++++- mm/gup.c | 27 ++++++- 6 files changed, 268 insertions(+), 10 deletions(-)