From: Pekka Enberg <[EMAIL PROTECTED]> As pointed out by Andrew Morton, if some process is partway through a big page_cache_readahead() operation, a concurrent invalidate_inode_pages2() is not enough to revoke inode pages. This patch changes the revoke code to modify ->f_mapping to point to a revoked mapping that will cause any future operations on the mapping to fail with EIO.
Signed-off-by: Pekka Enberg <[EMAIL PROTECTED]> --- fs/revoke.c | 29 ++++++++++++++++++++--------- fs/revoked_inode.c | 39 +++++++++++++++++++++++++++++++++++++++ include/linux/fs.h | 4 ++-- include/linux/mm.h | 1 + mm/mmap.c | 11 +++++++++++ 5 files changed, 73 insertions(+), 11 deletions(-) Index: uml-2.6/fs/revoke.c =================================================================== --- uml-2.6.orig/fs/revoke.c 2007-04-12 18:17:55.000000000 +0300 +++ uml-2.6/fs/revoke.c 2007-04-19 16:55:14.000000000 +0300 @@ -201,6 +201,10 @@ if (ret < 0) { err = ret; break; } + + unlink_file_vma(vma); + fput(vma->vm_file); + vma->vm_file = NULL; } up_read(&mm->mmap_sem); return err; @@ -290,6 +294,10 @@ int err = 0; err = revoke_vma(vma, &details); if (err) break; + + __unlink_file_vma(vma); + fput(vma->vm_file); + vma->vm_file = NULL; } up_write(&mm->mmap_sem); out: @@ -317,12 +325,10 @@ vma_prio_tree_foreach(vma, &iter, &mapp continue; err = revoke_mm(vma->vm_mm, mapping, to_exclude); - if (err == -EAGAIN) { + if (err == -EAGAIN) try_again = 1; - continue; - } - if (err == -EINTR) - goto restart; + + goto restart; } if (try_again) { cond_resched(); @@ -418,9 +424,11 @@ int err = 0; for (i = 0; i < table->end; i++) { struct revokefs_inode_info *info; struct file *this, *filp; + struct inode *inode; this = table->files[i]; - info = revokefs_i(this->f_dentry->d_inode); + inode = this->f_dentry->d_inode; + info = revokefs_i(inode); /* * Increase count before attempting to close file as @@ -428,7 +436,7 @@ for (i = 0; i < table->end; i++) { */ table->restore_start++; filp = info->file; - err = filp->f_op->revoke(filp); + err = filp->f_op->revoke(filp, inode->i_mapping); put_task_struct(info->owner); info->owner = NULL; /* To avoid restoring closed file. */ if (err) @@ -633,8 +641,9 @@ asmlinkage long sys_frevoke(unsigned int return err; } -int generic_file_revoke(struct file *file) +int generic_file_revoke(struct file *file, struct address_space *new_mapping) { + struct address_space *mapping = file->f_mapping; int err; /* @@ -644,10 +653,12 @@ int generic_file_revoke(struct file *fil if (err) goto out; + file->f_mapping = new_mapping; + /* * Make pending reads fail. */ - err = invalidate_inode_pages2(file->f_mapping); + err = invalidate_inode_pages2(mapping); out: return err; Index: uml-2.6/fs/revoked_inode.c =================================================================== --- uml-2.6.orig/fs/revoked_inode.c 2007-04-12 18:17:55.000000000 +0300 +++ uml-2.6/fs/revoked_inode.c 2007-04-12 18:29:30.000000000 +0300 @@ -362,6 +362,43 @@ static struct inode_operations revoked_i /* truncate_range returns void */ }; +static int revoked_readpage(struct file *file, struct page *page) +{ + return -EIO; +} + +static int revoked_writepage(struct page *page, struct writeback_control *wbc) +{ + return -EIO; +} + +static int revoked_prepare_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{ + return -EIO; +} + +static int revoked_commit_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{ + return -EIO; +} + +static ssize_t revoked_direct_IO(int rw, struct kiocb *iocb, + const struct iovec *iov, loff_t offset, + unsigned long nr_segs) +{ + return -EIO; +} + +static const struct address_space_operations revoked_aops = { + .readpage = revoked_readpage, + .writepage = revoked_writepage, + .prepare_write = revoked_prepare_write, + .commit_write = revoked_commit_write, + .direct_IO = revoked_direct_IO, +}; + void make_revoked_inode(struct inode *inode, int mode) { remove_inode_hash(inode); @@ -375,4 +412,6 @@ void make_revoked_inode(struct inode *in inode->i_fop = &revoked_special_file_ops; else inode->i_fop = &revoked_file_ops; + + inode->i_mapping->a_ops = &revoked_aops; } Index: uml-2.6/include/linux/fs.h =================================================================== --- uml-2.6.orig/include/linux/fs.h 2007-04-12 18:17:55.000000000 +0300 +++ uml-2.6/include/linux/fs.h 2007-04-12 18:29:30.000000000 +0300 @@ -1100,7 +1100,7 @@ struct file_operations { int (*flock) (struct file *, int, struct file_lock *); ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); - int (*revoke)(struct file *); + int (*revoke)(struct file *, struct address_space *); }; struct inode_operations { @@ -1742,7 +1742,7 @@ extern long do_splice_direct(struct file /* fs/revoke.c */ #ifdef CONFIG_MMU -extern int generic_file_revoke(struct file *); +extern int generic_file_revoke(struct file *, struct address_space *); #else #define generic_file_revoke NULL #endif Index: uml-2.6/include/linux/mm.h =================================================================== --- uml-2.6.orig/include/linux/mm.h 2007-04-12 16:57:33.000000000 +0300 +++ uml-2.6/include/linux/mm.h 2007-04-12 19:07:34.000000000 +0300 @@ -1022,6 +1022,7 @@ extern int split_vma(struct mm_struct *, extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, struct rb_node **, struct rb_node *); +extern void __unlink_file_vma(struct vm_area_struct *); extern void unlink_file_vma(struct vm_area_struct *); extern struct vm_area_struct *copy_vma(struct vm_area_struct **, unsigned long addr, unsigned long len, pgoff_t pgoff); Index: uml-2.6/mm/mmap.c =================================================================== --- uml-2.6.orig/mm/mmap.c 2007-04-12 16:57:33.000000000 +0300 +++ uml-2.6/mm/mmap.c 2007-04-19 16:55:47.000000000 +0300 @@ -201,6 +201,17 @@ static void __remove_shared_vm_struct(st } /* + * Requires inode->i_mapping->i_mmap_lock + */ +void __unlink_file_vma(struct vm_area_struct *vma) +{ + struct file *file = vma->vm_file; + struct address_space *mapping = file->f_mapping; + + __remove_shared_vm_struct(vma, file, mapping); +} + +/* * Unlink a file-based vm structure from its prio_tree, to hide * vma from rmap and vmtruncate before freeing its page tables. */ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/