Attachment 313588 Details for Bug 455408 – Draft Proposed patch

[patch] Draft Proposed patch

455408.rhel-5.2 (text/plain), 9.09 KB, created by Peter Staubach on 2008-08-06 15:40:01 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Peter Staubach

Created: 2008-08-06 15:40:01 UTC

Size: 9.09 KB

patch

obsolete

>--- linux-2.6.18.i686/fs/nfs/file.c.org
>+++ linux-2.6.18.i686/fs/nfs/file.c
>@@ -390,6 +390,9 @@ nfs_file_write(struct kiocb *iocb, const
> 	result = -EBUSY;
> 	if (IS_SWAPFILE(inode))
> 		goto out_swapfile;
>+
>+	nfs_writes_outstanding(NFS_SERVER(inode));
>+
> 	/*
> 	 * O_APPEND implies that we must revalidate the file length.
> 	 */
>@@ -405,11 +408,17 @@ nfs_file_write(struct kiocb *iocb, const
> 
> 	nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
> 	result = generic_file_aio_write(iocb, buf, count, pos);
>-	/* Return error values for O_SYNC and IS_SYNC() */
>-	if (result >= 0 && (IS_SYNC(inode) || (iocb->ki_filp->f_flags & O_SYNC))) {
>-		int err = nfs_fsync(iocb->ki_filp, dentry, 1);
>-		if (err < 0)
>-			result = err;
>+	if (result >= 0) {
>+		if (IS_SYNC(inode) || (iocb->ki_filp->f_flags & O_SYNC)) {
>+			/* Return error values for O_SYNC and IS_SYNC() */
>+			int err = nfs_fsync(iocb->ki_filp, dentry, 1);
>+			if (err < 0)
>+				result = err;
>+		} else if (((pos + count) & (NFS_SERVER(inode)->wsize - 1)) == 0) {
>+			nfs_sync_inode_nowait(inode,
>+				(pos & ~(NFS_SERVER(inode)->wsize - 1)) >> PAGE_CACHE_SHIFT,
>+				NFS_SERVER(inode)->wpages);
>+		}
> 	}
> out:
> 	return result;
>--- linux-2.6.18.i686/fs/nfs/write.c.org
>+++ linux-2.6.18.i686/fs/nfs/write.c
>@@ -1061,6 +1061,7 @@ static int nfs_flush_multi(struct inode 
> 					nbytes, offset, how);
> 			nbytes = 0;
> 		}
>+		nfs_writes_active(NFS_SERVER(inode));
> 		nfs_execute_write(data);
> 	} while (nbytes != 0);
> 
>@@ -1112,6 +1113,7 @@ static int nfs_flush_one(struct inode *i
> 	/* Set up the argument struct */
> 	nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
> 
>+	nfs_writes_active(NFS_SERVER(inode));
> 	nfs_execute_write(data);
> 	return 0;
>  out_bad:
>@@ -1136,10 +1138,12 @@ static int nfs_flush_list(struct inode *
> 	flush_one = nfs_flush_one;
> 	if (wsize < PAGE_CACHE_SIZE)
> 		flush_one = nfs_flush_multi;
>+#ifdef notdef
> 	/* For single writes, FLUSH_STABLE is more efficient */
> 	if (npages <= wpages && npages == NFS_I(inode)->npages
> 			&& nfs_list_entry(head->next)->wb_bytes <= wsize)
> 		how |= FLUSH_STABLE;
>+#endif
> 
> 	do {
> 		nfs_coalesce_requests(head, &one_request, wpages);
>@@ -1340,6 +1344,7 @@ int nfs_writeback_done(struct rpc_task *
> 		/* Can't do anything about it except throw an error. */
> 		task->tk_status = -EIO;
> 	}
>+	nfs_writes_done(NFS_SERVER(data->inode));
> 	return 0;
> }
> 
>@@ -1561,6 +1566,25 @@ int nfs_sync_inode_wait(struct inode *in
> 	return ret;
> }
> 
>+void nfs_sync_inode_nowait(struct inode *inode, unsigned long idx_start,
>+		unsigned int npages)
>+{
>+	struct nfs_inode *nfsi = NFS_I(inode);
>+	LIST_HEAD(head);
>+	int pages, ret = 0;
>+
>+	spin_lock(&nfsi->req_lock);
>+	do {
>+		pages = nfs_scan_dirty(inode, &head, idx_start, npages);
>+		if (pages != 0) {
>+			spin_unlock(&nfsi->req_lock);
>+			ret = nfs_flush_list(inode, &head, pages, 0);
>+			spin_lock(&nfsi->req_lock);
>+		}
>+	} while (pages && !ret);
>+	spin_unlock(&nfsi->req_lock);
>+}
>+
> int __init nfs_init_writepagecache(void)
> {
> 	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
>--- linux-2.6.18.i686/fs/nfs/sysctl.c.org
>+++ linux-2.6.18.i686/fs/nfs/sysctl.c
>@@ -98,6 +98,22 @@ static ctl_table nfs_cb_sysctls[] = {
> 		.proc_handler = &proc_dointvec,
> 	},
> #endif
>+	{
>+		.ctl_name = CTL_UNNUMBERED,
>+		.procname = "nfs_max_outstanding_writes",
>+		.data = &nfs_max_outstanding_writes,
>+		.maxlen = sizeof(int),
>+		.mode = 0644,
>+		.proc_handler = &proc_dointvec,
>+	},
>+	{
>+		.ctl_name = CTL_UNNUMBERED,
>+		.procname = "nfs_peak_outstanding_writes",
>+		.data = &nfs_peak_outstanding_writes,
>+		.maxlen = sizeof(int),
>+		.mode = 0644,
>+		.proc_handler = &proc_dointvec,
>+	},
> 	{ .ctl_name = 0 }
> };
> 
>--- linux-2.6.18.i686/fs/nfs/client.c.org
>+++ linux-2.6.18.i686/fs/nfs/client.c
>@@ -770,6 +770,9 @@ static struct nfs_server *nfs_alloc_serv
> 	init_waitqueue_head(&server->active_wq);
> 	atomic_set(&server->active, 0);
> 
>+	init_waitqueue_head(&server->writes_wq);
>+	atomic_set(&server->writes, 0);
>+
> 	server->io_stats = nfs_alloc_iostats();
> 	if (!server->io_stats) {
> 		kfree(server);
>--- linux-2.6.18.i686/fs/nfs/super.c.org
>+++ linux-2.6.18.i686/fs/nfs/super.c
>@@ -209,6 +209,32 @@ static void nfs_put_super(struct super_b
> 	wait_event(server->active_wq, atomic_read(&server->active) == 0);
> }
> 
>+int nfs_max_outstanding_writes = 0;
>+int nfs_peak_outstanding_writes = 0;
>+
>+void nfs_writes_active(struct nfs_server *server)
>+{
>+	int peak;
>+
>+	peak = atomic_inc_return(&server->writes);
>+	if (peak > nfs_peak_outstanding_writes)
>+		nfs_peak_outstanding_writes = peak;
>+}
>+
>+void nfs_writes_done(struct nfs_server *server)
>+{
>+	if (atomic_dec_return(&server->writes) < nfs_max_outstanding_writes / 2)
>+		wake_up(&server->writes_wq);
>+}
>+
>+void nfs_writes_outstanding(struct nfs_server *server)
>+{
>+	if (nfs_max_outstanding_writes) {
>+		while (atomic_read(&server->writes) >= nfs_max_outstanding_writes)
>+			wait_event(server->writes_wq, atomic_read(&server->writes) < nfs_max_outstanding_writes / 2);
>+	}
>+}
>+
> /*
>  * Deliver file system statistics to userspace
>  */
>--- linux-2.6.18.i686/fs/nfs/internal.h.org
>+++ linux-2.6.18.i686/fs/nfs/internal.h
>@@ -168,6 +168,12 @@ extern void __exit unregister_nfs_fs(voi
> extern void nfs_sb_active(struct nfs_server *server);
> extern void nfs_sb_deactive(struct nfs_server *server);
> 
>+extern void nfs_writes_active(struct nfs_server *server);
>+extern void nfs_writes_done(struct nfs_server *server);
>+extern void nfs_writes_outstanding(struct nfs_server *server);
>+extern int nfs_max_outstanding_writes;
>+extern int nfs_peak_outstanding_writes;
>+
> /* namespace.c */
> extern char *nfs_path(const char *base,
> 		      const struct dentry *droot,
>--- linux-2.6.18.i686/include/linux/nfs_fs.h.org
>+++ linux-2.6.18.i686/include/linux/nfs_fs.h
>@@ -450,6 +450,7 @@ extern void nfs_writedata_release(void *
>  * return value!)
>  */
> extern int  nfs_sync_inode_wait(struct inode *, unsigned long, unsigned int, int);
>+extern void nfs_sync_inode_nowait(struct inode *, unsigned long, unsigned int);
> #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
> extern int  nfs_commit_inode(struct inode *, int);
> extern struct nfs_write_data *nfs_commit_alloc(void);
>--- linux-2.6.18.i686/include/linux/sysctl.h.org
>+++ linux-2.6.18.i686/include/linux/sysctl.h
>@@ -205,6 +205,7 @@ enum
> 	VM_PAGECACHE=37,        /* favor reclaiming unmapped pagecache pages */
> 	VM_MMAP_MIN_ADDR=38, 	/* amount of memory to protect from mmap */
> 	VM_FLUSH_MMAP=39,       /* flush mmap()d pagecache pages */
>+	VM_MAX_QUEUE_DEPTH=40,	/* max IO queue dpeth */
> };
> 
> 
>--- linux-2.6.18.i686/include/linux/writeback.h.org
>+++ linux-2.6.18.i686/include/linux/writeback.h
>@@ -89,6 +89,7 @@ void throttle_vm_writeout(void);
> /* These are exported to sysctl. */
> extern int dirty_background_ratio;
> extern int vm_dirty_ratio;
>+extern int vm_max_queue_depth;
> extern int dirty_writeback_interval;
> extern int dirty_expire_interval;
> extern int block_dump;
>--- linux-2.6.18.i686/include/linux/nfs_fs_sb.h.org
>+++ linux-2.6.18.i686/include/linux/nfs_fs_sb.h
>@@ -122,6 +122,8 @@ struct nfs_server {
> 
> 	atomic_t active; /* Keep trace of any activity to this server */
> 	wait_queue_head_t active_wq;  /* Wait for any activity to stop  */
>+	atomic_t writes;	/* number of outstanding WRITE requests */
>+	wait_queue_head_t writes_wq;
> };
> 
> /* Server capabilities */
>--- linux-2.6.18.i686/kernel/sysctl.c.org
>+++ linux-2.6.18.i686/kernel/sysctl.c
>@@ -1130,6 +1130,17 @@ static ctl_table vm_table[] = {
> 		.strategy	= &sysctl_intvec,
> 		.extra1		= &zero,
> 	},
>+	{
>+		.ctl_name       = VM_MAX_QUEUE_DEPTH,
>+		.procname       = "max_queue_depth",
>+		.data           = &vm_max_queue_depth,
>+		.maxlen         = sizeof(vm_max_queue_depth),
>+		.mode           = 0644,
>+		.proc_handler   = &proc_dointvec_minmax,
>+		.strategy       = &sysctl_intvec,
>+		.extra1         = &zero,
>+		.extra2         = &one_hundred,
>+	},
> 	{ .ctl_name = 0 }
> };
> 
>--- linux-2.6.18.i686/mm/page-writeback.c.org
>+++ linux-2.6.18.i686/mm/page-writeback.c
>@@ -72,6 +72,8 @@ int dirty_background_ratio = 10;
>  */
> int vm_dirty_ratio = 40;
> 
>+int vm_max_queue_depth = 0;
>+
> /*
>  * The interval between `kupdate'-style writebacks, in jiffies
>  */
>@@ -147,12 +149,12 @@ get_dirty_limits(long *pbackground, long
> 	if (dirty_ratio > unmapped_ratio / 2)
> 		dirty_ratio = unmapped_ratio / 2;
> 
>-	if (dirty_ratio < 5)
>-		dirty_ratio = 5;
>+	if (dirty_ratio < 1)
>+		dirty_ratio = 1;
> 
> 	background_ratio = dirty_background_ratio;
> 	if (background_ratio >= dirty_ratio)
>-		background_ratio = dirty_ratio / 2;
>+		background_ratio = ((dirty_ratio>1)?dirty_ratio/2:dirty_ratio);
> 
> 	background = (background_ratio * available_memory) / 100;
> 	dirty = (dirty_ratio * available_memory) / 100;
>@@ -161,6 +163,18 @@ get_dirty_limits(long *pbackground, long
> 		background += background / 4;
> 		dirty += dirty / 4;
> 	}
>+	/* vm_max_queue_depth allows setting the dirty_ratio to less than 1% of RAM
>+	 * and prevents the dirty_background_ratio from being greater than dirty_ratio.
>+	 */
>+	if (vm_max_queue_depth && (vm_max_queue_depth<<(20-PAGE_SHIFT)) < dirty) {
>+		dirty = vm_max_queue_depth<<(20-PAGE_SHIFT);
>+		if (background > dirty) {
>+			if (dirty_ratio > background_ratio)
>+				background = dirty/2;
>+			else
>+				background = dirty;
>+		}
>+	}
> 	*pbackground = background;
> 	*pdirty = dirty;
> }

Actions: View | Diff

Attachments on bug 455408: 311973 | 312656 | 313588 | 314535 | 316301 | 316870 | 317393