>>>>> "juan" == Juan J Quintela <[EMAIL PROTECTED]> writes:
Hi Linus
I am resending to you this patch, it works here, and nobody has
complained about it the previous times that I post it here.
Later, Juan.
juan> I have reworte the function truncate_inode_pages. The version
juan> in vanilla pre9-3 does busy waiting in the partial page, with
juan> this version the locking for the partial page and from the
juan> rest of the pages is the same. This make that we have less
juan> special cases. For the rest of pages the function works the
juan> same. The only difference is that version is cleaner IMHO.
juan> Or there are some corner case that I have failed to see?
juan> Comments?
juan> Later, Juan.
juan> I have CC: the linux-fsdevel people, they are the users of
juan> that function, could somebody give me some feedback against
juan> the change?
diff -urN --exclude-from=/home/lfcia/quintela/work/kernel/exclude work/mm/filemap.c
testing/mm/filemap.c
--- work/mm/filemap.c Fri May 12 23:46:46 2000
+++ testing/mm/filemap.c Sun May 14 22:08:45 2000
@@ -146,9 +146,39 @@
spin_unlock(&pagecache_lock);
}
-/*
+static inline void truncate_partial_page(struct page *page, unsigned partial)
+{
+ memclear_highpage_flush(page, partial, PAGE_CACHE_SIZE-partial);
+
+ if (page->buffers)
+ block_flushpage(page, partial);
+
+}
+
+static inline void truncate_complete_page(struct page *page)
+{
+ if (!page->buffers || block_flushpage(page, 0))
+ lru_cache_del(page);
+
+ /*
+ * We remove the page from the page cache _after_ we have
+ * destroyed all buffer-cache references to it. Otherwise some
+ * other process might think this inode page is not in the
+ * page cache and creates a buffer-cache alias to it causing
+ * all sorts of fun problems ...
+ */
+ remove_inode_page(page);
+ page_cache_release(page);
+}
+
+/**
+ * truncate_inode_pages - truncate *all* the pages from an offset
+ * @mapping: mapping to truncate
+ * @lstart: offset from with to truncate
+ *
* Truncate the page cache at a set offset, removing the pages
* that are beyond that offset (and zeroing out partial pages).
+ * If any page is locked we wait for it to become unlocked.
*/
void truncate_inode_pages(struct address_space * mapping, loff_t lstart)
{
@@ -168,11 +198,10 @@
page = list_entry(curr, struct page, list);
curr = curr->next;
-
offset = page->index;
- /* page wholly truncated - free it */
- if (offset >= start) {
+ /* Is one of the pages to truncate? */
+ if ((offset >= start) || (partial && (offset + 1) == start)) {
if (TryLockPage(page)) {
page_cache_get(page);
spin_unlock(&pagecache_lock);
@@ -183,22 +212,14 @@
page_cache_get(page);
spin_unlock(&pagecache_lock);
- if (!page->buffers || block_flushpage(page, 0))
- lru_cache_del(page);
-
- /*
- * We remove the page from the page cache
- * _after_ we have destroyed all buffer-cache
- * references to it. Otherwise some other process
- * might think this inode page is not in the
- * page cache and creates a buffer-cache alias
- * to it causing all sorts of fun problems ...
- */
- remove_inode_page(page);
+ if (partial && (offset + 1) == start) {
+ truncate_partial_page(page, partial);
+ partial = 0;
+ } else
+ truncate_complete_page(page);
UnlockPage(page);
page_cache_release(page);
- page_cache_release(page);
/*
* We have done things without the pagecache lock,
@@ -209,37 +230,6 @@
*/
goto repeat;
}
- /*
- * there is only one partial page possible.
- */
- if (!partial)
- continue;
-
- /* and it's the one preceeding the first wholly truncated page */
- if ((offset + 1) != start)
- continue;
-
- /* partial truncate, clear end of page */
- if (TryLockPage(page)) {
- spin_unlock(&pagecache_lock);
- goto repeat;
- }
- page_cache_get(page);
- spin_unlock(&pagecache_lock);
-
- memclear_highpage_flush(page, partial, PAGE_CACHE_SIZE-partial);
- if (page->buffers)
- block_flushpage(page, partial);
-
- partial = 0;
-
- /*
- * we have dropped the spinlock so we have to
- * restart.
- */
- UnlockPage(page);
- page_cache_release(page);
- goto repeat;
}
spin_unlock(&pagecache_lock);
}
--
In theory, practice and theory are the same, but in practice they
are different -- Larry McVoy