]> bbs.cooldavid.org Git - net-next-2.6.git/blobdiff - mm/filemap.c
[PATCH] BLOCK: Make it possible to disable the block layer [try #6]
[net-next-2.6.git] / mm / filemap.c
index 9c7334bafda8da5150f4a9f0a7556bda113b8a19..c4fe97f5ace0851d3023d1305d8de5c7ad2b1dd5 100644 (file)
@@ -9,7 +9,6 @@
  * most "normal" filesystems (but you don't /have/ to use this:
  * the NFS filesystem used to do this differently, for example)
  */
-#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/compiler.h>
@@ -120,7 +119,7 @@ void __remove_from_page_cache(struct page *page)
        radix_tree_delete(&mapping->page_tree, page->index);
        page->mapping = NULL;
        mapping->nrpages--;
-       pagecache_acct(-1);
+       __dec_zone_page_state(page, NR_FILE_PAGES);
 }
 
 void remove_from_page_cache(struct page *page)
@@ -449,7 +448,7 @@ int add_to_page_cache(struct page *page, struct address_space *mapping,
                        page->mapping = mapping;
                        page->index = offset;
                        mapping->nrpages++;
-                       pagecache_acct(1);
+                       __inc_zone_page_state(page, NR_FILE_PAGES);
                }
                write_unlock_irq(&mapping->tree_lock);
                radix_tree_preload_end();
@@ -489,6 +488,12 @@ struct page *page_cache_alloc_cold(struct address_space *x)
 EXPORT_SYMBOL(page_cache_alloc_cold);
 #endif
 
+static int __sleep_on_page_lock(void *word)
+{
+       io_schedule();
+       return 0;
+}
+
 /*
  * In order to wait for pages to become available there must be
  * waitqueues associated with pages. By using a hash table of
@@ -578,13 +583,24 @@ void fastcall __lock_page(struct page *page)
 }
 EXPORT_SYMBOL(__lock_page);
 
+/*
+ * Variant of lock_page that does not require the caller to hold a reference
+ * on the page's mapping.
+ */
+void fastcall __lock_page_nosync(struct page *page)
+{
+       DEFINE_WAIT_BIT(wait, &page->flags, PG_locked);
+       __wait_on_bit_lock(page_waitqueue(page), &wait, __sleep_on_page_lock,
+                                                       TASK_UNINTERRUPTIBLE);
+}
+
 /**
  * find_get_page - find and get a page reference
  * @mapping: the address_space to search
  * @offset: the page index
  *
- * A rather lightweight function, finding and getting a reference to a
- * hashed page atomically.
+ * Is there a pagecache struct page at the given (mapping, offset) tuple?
+ * If yes, increment its refcount and return it; if no, return NULL.
  */
 struct page * find_get_page(struct address_space *mapping, unsigned long offset)
 {
@@ -850,8 +866,6 @@ static void shrink_readahead_size_eio(struct file *filp,
                return;
 
        ra->ra_pages /= 4;
-       printk(KERN_WARNING "Reducing readahead size to %luK\n",
-                       ra->ra_pages << (PAGE_CACHE_SHIFT - 10));
 }
 
 /**
@@ -973,7 +987,7 @@ page_not_up_to_date:
                /* Get exclusive access to the page ... */
                lock_page(page);
 
-               /* Did it get unhashed before we got the lock? */
+               /* Did it get truncated before we got the lock? */
                if (!page->mapping) {
                        unlock_page(page);
                        page_cache_release(page);
@@ -1416,7 +1430,7 @@ retry_find:
                 */
                if (!did_readaround) {
                        majmin = VM_FAULT_MAJOR;
-                       inc_page_state(pgmajfault);
+                       count_vm_event(PGMAJFAULT);
                }
                did_readaround = 1;
                ra_pages = max_sane_readahead(file->f_ra.ra_pages);
@@ -1457,7 +1471,7 @@ outside_data_content:
         * accessible..
         */
        if (area->vm_mm == current->mm)
-               return NULL;
+               return NOPAGE_SIGBUS;
        /* Fall through to the non-read-ahead case */
 no_cached_page:
        /*
@@ -1482,12 +1496,12 @@ no_cached_page:
         */
        if (error == -ENOMEM)
                return NOPAGE_OOM;
-       return NULL;
+       return NOPAGE_SIGBUS;
 
 page_not_uptodate:
        if (!did_readaround) {
                majmin = VM_FAULT_MAJOR;
-               inc_page_state(pgmajfault);
+               count_vm_event(PGMAJFAULT);
        }
        lock_page(page);
 
@@ -1551,7 +1565,7 @@ page_not_uptodate:
         */
        shrink_readahead_size_eio(file, ra);
        page_cache_release(page);
-       return NULL;
+       return NOPAGE_SIGBUS;
 }
 EXPORT_SYMBOL(filemap_nopage);
 
@@ -1613,7 +1627,7 @@ no_cached_page:
 page_not_uptodate:
        lock_page(page);
 
-       /* Did it get unhashed while we waited for it? */
+       /* Did it get truncated while we waited for it? */
        if (!page->mapping) {
                unlock_page(page);
                goto err;
@@ -2006,6 +2020,7 @@ inline int generic_write_checks(struct file *file, loff_t *pos, size_t *count, i
                if (unlikely(*pos + *count > inode->i_sb->s_maxbytes))
                        *count = inode->i_sb->s_maxbytes - *pos;
        } else {
+#ifdef CONFIG_BLOCK
                loff_t isize;
                if (bdev_read_only(I_BDEV(inode)))
                        return -EPERM;
@@ -2017,6 +2032,9 @@ inline int generic_write_checks(struct file *file, loff_t *pos, size_t *count, i
 
                if (*pos + *count > isize)
                        *count = isize - *pos;
+#else
+               return -EPERM;
+#endif
        }
        return 0;
 }
@@ -2069,7 +2087,7 @@ generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
 {
        struct file *file = iocb->ki_filp;
        struct address_space * mapping = file->f_mapping;
-       struct address_space_operations *a_ops = mapping->a_ops;
+       const struct address_space_operations *a_ops = mapping->a_ops;
        struct inode    *inode = mapping->host;
        long            status = 0;
        struct page     *page;
@@ -2095,14 +2113,21 @@ generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
        do {
                unsigned long index;
                unsigned long offset;
-               unsigned long maxlen;
                size_t copied;
 
                offset = (pos & (PAGE_CACHE_SIZE -1)); /* Within page */
                index = pos >> PAGE_CACHE_SHIFT;
                bytes = PAGE_CACHE_SIZE - offset;
-               if (bytes > count)
-                       bytes = count;
+
+               /* Limit the size of the copy to the caller's write size */
+               bytes = min(bytes, count);
+
+               /*
+                * Limit the size of the copy to that of the current segment,
+                * because fault_in_pages_readable() doesn't know how to walk
+                * segments.
+                */
+               bytes = min(bytes, cur_iov->iov_len - iov_base);
 
                /*
                 * Bring in the user page that we will copy from _first_.
@@ -2110,10 +2135,7 @@ generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
                 * same page as we're writing to, without it being marked
                 * up-to-date.
                 */
-               maxlen = cur_iov->iov_len - iov_base;
-               if (maxlen > bytes)
-                       maxlen = bytes;
-               fault_in_pages_readable(buf, maxlen);
+               fault_in_pages_readable(buf, bytes);
 
                page = __grab_cache_page(mapping,index,&cached_page,&lru_pvec);
                if (!page) {
@@ -2121,6 +2143,12 @@ generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
                        break;
                }
 
+               if (unlikely(bytes == 0)) {
+                       status = 0;
+                       copied = 0;
+                       goto zero_length_segment;
+               }
+
                status = a_ops->prepare_write(file, page, offset, offset+bytes);
                if (unlikely(status)) {
                        loff_t isize = i_size_read(inode);
@@ -2150,7 +2178,8 @@ generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
                        page_cache_release(page);
                        continue;
                }
-               if (likely(copied > 0)) {
+zero_length_segment:
+               if (likely(copied >= 0)) {
                        if (!status)
                                status = copied;
 
@@ -2215,7 +2244,7 @@ __generic_file_aio_write_nolock(struct kiocb *iocb, const struct iovec *iov,
                                unsigned long nr_segs, loff_t *ppos)
 {
        struct file *file = iocb->ki_filp;
-       struct address_space * mapping = file->f_mapping;
+       const struct address_space * mapping = file->f_mapping;
        size_t ocount;          /* original count */
        size_t count;           /* after file limit checks */
        struct inode    *inode = mapping->host;
@@ -2466,3 +2495,33 @@ generic_file_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
        }
        return retval;
 }
+
+/**
+ * try_to_release_page() - release old fs-specific metadata on a page
+ *
+ * @page: the page which the kernel is trying to free
+ * @gfp_mask: memory allocation flags (and I/O mode)
+ *
+ * The address_space is to try to release any data against the page
+ * (presumably at page->private).  If the release was successful, return `1'.
+ * Otherwise return zero.
+ *
+ * The @gfp_mask argument specifies whether I/O may be performed to release
+ * this page (__GFP_IO), and whether the call may block (__GFP_WAIT).
+ *
+ * NOTE: @gfp_mask may go away, and this function may become non-blocking.
+ */
+int try_to_release_page(struct page *page, gfp_t gfp_mask)
+{
+       struct address_space * const mapping = page->mapping;
+
+       BUG_ON(!PageLocked(page));
+       if (PageWriteback(page))
+               return 0;
+
+       if (mapping && mapping->a_ops->releasepage)
+               return mapping->a_ops->releasepage(page, gfp_mask);
+       return try_to_free_buffers(page);
+}
+
+EXPORT_SYMBOL(try_to_release_page);