Skip to contentSkip to conversation

How to completely disable async fastboot?

DISCUSSION Jeff Garzik 3 participants 4 messages Apr 16, 2009 – Apr 17, 2009

lore
Share & tools
lore
#0 lore raw

The following commit

Quoted 5 lines

commit 9710794383ee5008d67f1a6613a4717bf6de47bc

Author: Arjan van de Ven <[email protected]>
Date:   Sun Mar 15 11:11:44 2009 -0700

async: remove the temporary (2.6.29) "async is off by default" code

not only changes a default, it completely removes the ability to turn off async AFAICS.

At present, I need to turn this off to debug boot-time operation.

Jeff

#1 lore raw
Quoted 11 lines · On Wed, 15 Apr 2009 20:20:32 -0400 Jeff Garzik <[email protected]> wrote:

The following commit

Quoted 6 lines

commit 9710794383ee5008d67f1a6613a4717bf6de47bc

Author: Arjan van de Ven <[email protected]>
Date:   Sun Mar 15 11:11:44 2009 -0700

async: remove the temporary (2.6.29) "async is off by default"
code

not only changes a default, it completely removes the ability to turn off async AFAICS.

this is on the request of Andrew and others who want to be able to use these API's (soon) in an "forced offload" scenario.

Quoted 2 lines

At present, I need to turn this off to debug boot-time operation.

can you give some details? Maybe I can help suggest things...

Signature · 3 lines
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
#2 lore raw
Quoted 11 lines · Arjan van de Ven
Quoted 8 lines · On Wed, 15 Apr 2009 20:20:32 -0400 Jeff Garzik <[email protected]> wrote:

The following commit

Quoted 6 lines

commit 9710794383ee5008d67f1a6613a4717bf6de47bc

Author: Arjan van de Ven <[email protected]>
Date:   Sun Mar 15 11:11:44 2009 -0700

async: remove the temporary (2.6.29) "async is off by default"
code

Quoted 8 lines
Quoted 1 line

not only changes a default, it completely removes the ability to turn off async AFAICS.

this is on the request of Andrew and others who want to be able to use these API's (soon) in an "forced offload" scenario.

Quoted 1 line

At present, I need to turn this off to debug boot-time operation.

can you give some details? Maybe I can help suggest things...

Too much shit is occurring in parallel, which completely obscures the operation of the driver I am currently debugging. A new and unwelcome behavior, as it is directly impacting workflow.

Commit 9710794383ee5008d67f1a6613a4717bf6de47bc was fine for the normal case, but please have pity on developers! Just like many multi-threaded userland programs have an option to run single-threaded (usually for debug purposes), the kernel should have this too.

I need a slowboot option, like the attached patch.

The patch is basically the reverse of your 9710794383ee5008d67f1a6613a4717bf6de47bc in a different way.

Jeff

kernel/async.c +14 -0
index 968ef94..8c553ce 100644
@@ -68,6 +68,7 @@ static LIST_HEAD(async_running);
 static DEFINE_SPINLOCK(async_lock);
 
 static int async_enabled = 0;
+static int async_disabled;
 
 struct async_entry {
 	struct list_head list;
@@ -388,6 +389,11 @@ static int async_manager_thread(void *unused)
 
 static int __init async_init(void)
 {
+	if (async_disabled) {
+		async_enabled = 0;
+		return 0;
+	}
+
 	async_enabled =
 		!IS_ERR(kthread_run(async_manager_thread, NULL, "async/mgr"));
 
@@ -395,4 +401,12 @@ static int __init async_init(void)
 	return 0;
 }
 
+static int __init setup_async(char *str)
+{
+	async_disabled = 1;
+	return 1;
+}
+
+__setup("slowboot", setup_async);
+
 core_initcall(async_init);
#3 lore raw
In reply to Arjan van de Ven
Quoted 3 lines · Arjan van de Ven
Quoted 1 line · On Wed, 15 Apr 2009 20:20:32 -0400 Jeff Garzik <[email protected]> wrote:
Quoted 1 line

commit 9710794383ee5008d67f1a6613a4717bf6de47bc

...

Quoted 5 lines
Quoted 2 lines

not only changes a default, it completely removes the ability to turn off async AFAICS.

this is on the request of Andrew and others who want to be able to use these API's (soon) in an "forced offload" scenario.

Indeed, because the async API is a threadpool API.
However, 2.6.30-rc1 also gained David Howell's <linux/slow-work.h>, and I for one am still wondering which one of these I'm going to use...

Signature · 3 lines
Stefan Richter
-=====-==--= -=-- =---=
http://arcgraph.de/sr/