Re: [PATCH v2] Make vm_max_readahead configurable at run-time

[Nikanth Karthikesan <knikanth@suse.de>]

On Thursday 11 February 2010 10:43:41 Wu Fengguang wrote:
> On Wed, Feb 10, 2010 at 09:52:40PM +0800, Nikanth Karthikesan wrote:
> > On Wednesday 10 February 2010 16:35:51 Wu Fengguang wrote:
> > > Nikanth,
> > >
> > > > Make vm_max_readahead configurable at run-time. Expose a sysctl knob
> > > > in procfs to change it. This would ensure that new disks added would
> > > > use this value as their default read_ahead_kb.
> > >
> > > Do you have use case, or customer demand for it?
> >
> > No body requested for it. But when doing some performance testing with
> > readahead_kb re-compiling would be a pain, and thought that having a
> > configurable default might be useful.
> 
> I wonder why you need to recompile kernel in the tests.
> There are three interfaces to change readahead size in runtime:
> 
>         blockdev --setra 1024 /dev/sda
>         echo 512 > /sys/block/*/queue/read_ahead_kb
>         echo 512 > /sys/devices/virtual/bdi/*/read_ahead_kb
> 

Right, I did use that. But thought that a global tunable would be better. :-)

> > > > Also filesystems which use default_backing_dev_info would also
> > > > use this new value, even if they were already mounted.
> > > >
> > > > Currently xfs, btrfs, nilfs, raw, mtd use the
> > > > default_backing_dev_info.
> > >
> > > This sounds like bad interface, in that users will be confused by the
> > > tricky details of "works for new devices" and "works for some fs".
> > >
> > > One more tricky point is, btrfs/md/dm readahead size may not be
> > > influenced if some of the component disks are hot added.
> > >
> > > So this patch is only going to work for hot-plugged disks that
> > > contains _standalone_ filesystem. Is this typical use case in servers?
> >
> > Yes, it would work only if the top-level disk is hot-plugged/created.
> 
> Or maybe what you really want is a kernel parameter for setting the
> default readahead size at boot time?
> 
> In another thread, Christian Ehrhardt recommended to add a config
> option for it. If you like it, I can also do the kernel parameter
> by the way.
> 

Kernel parameter seems to be the right way to go. Does the attached patch look
good?

Thanks
Nikanth

From: Nikanth Karthikesan <knikanth@suse.de>

Add new kernel parameter "readahead", which would be used
as the value of VM_MAX_READAHEAD.

Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>

---

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 736d456..354e6f1 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -2148,6 +2148,8 @@ and is between 256 and 4096 characters. It is defined in the file
 			Format: <reboot_mode>[,<reboot_mode2>[,...]]
 			See arch/*/kernel/reboot.c or arch/*/kernel/process.c
 
+	readahead=	Default readahead value for block devices.
+
 	relax_domain_level=
 			[KNL, SMP] Set scheduler's default relax_domain_level.
 			See Documentation/cgroups/cpusets.txt.
diff --git a/block/blk-core.c b/block/blk-core.c
index 718897e..02ed748 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -499,7 +499,7 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 	q->backing_dev_info.unplug_io_fn = blk_backing_dev_unplug;
 	q->backing_dev_info.unplug_io_data = q;
 	q->backing_dev_info.ra_pages =
-			(VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE;
+			(vm_max_readahead_kb * 1024) / PAGE_CACHE_SIZE;
 	q->backing_dev_info.state = 0;
 	q->backing_dev_info.capabilities = BDI_CAP_MAP_COPY;
 	q->backing_dev_info.name = "block";
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 1a822ce..a593578 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -870,7 +870,7 @@ static int fuse_bdi_init(struct fuse_conn *fc, struct super_block *sb)
 	int err;
 
 	fc->bdi.name = "fuse";
-	fc->bdi.ra_pages = (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE;
+	fc->bdi.ra_pages = (vm_max_readahead_kb * 1024) / PAGE_CACHE_SIZE;
 	fc->bdi.unplug_io_fn = default_unplug_io_fn;
 	/* fuse does it's own writeback accounting */
 	fc->bdi.capabilities = BDI_CAP_NO_ACCT_WB;
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 60c467b..17825d7 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1188,9 +1188,11 @@ int write_one_page(struct page *page, int wait);
 void task_dirty_inc(struct task_struct *tsk);
 
 /* readahead.c */
-#define VM_MAX_READAHEAD	128	/* kbytes */
+#define DEFAULT_VM_MAX_READAHEAD       128     /* kbytes */
 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
 
+extern unsigned long vm_max_readahead_kb;
+
 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
 			pgoff_t offset, unsigned long nr_to_read);
 
diff --git a/init/main.c b/init/main.c
index 4cb47a1..7d5230a 100644
--- a/init/main.c
+++ b/init/main.c
@@ -70,6 +70,7 @@
 #include <linux/sfi.h>
 #include <linux/shmem_fs.h>
 #include <trace/boot.h>
+#include <linux/backing-dev.h>
 
 #include <asm/io.h>
 #include <asm/bugs.h>
@@ -249,6 +250,16 @@ static int __init loglevel(char *str)
 
 early_param("loglevel", loglevel);
 
+static int __init readahead(char *str)
+{
+	vm_max_readahead_kb = memparse(str, &str) / 1024ULL;
+	default_backing_dev_info.ra_pages = vm_max_readahead_kb
+						* 1024 / PAGE_CACHE_SIZE;
+	return 0;
+}
+
+early_param("readahead", readahead);
+
 /*
  * Unknown boot options get handed to init, unless they look like
  * unused parameters (modprobe will find them in /proc/cmdline).
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 0e8ca03..e33ff34 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -18,7 +18,7 @@ EXPORT_SYMBOL(default_unplug_io_fn);
 
 struct backing_dev_info default_backing_dev_info = {
 	.name		= "default",
-	.ra_pages	= VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE,
+	.ra_pages	= DEFAULT_VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE,
 	.state		= 0,
 	.capabilities	= BDI_CAP_MAP_COPY,
 	.unplug_io_fn	= default_unplug_io_fn,
diff --git a/mm/readahead.c b/mm/readahead.c
index 033bc13..516f8da 100644
--- a/mm/readahead.c
+++ b/mm/readahead.c
@@ -17,6 +17,8 @@
 #include <linux/pagevec.h>
 #include <linux/pagemap.h>
 
+unsigned long vm_max_readahead_kb = DEFAULT_VM_MAX_READAHEAD;
+
 /*
  * Initialise a struct file's readahead state.  Assumes that the caller has
  * memset *ra to zero.