[PATCH] RFC - add prealloc_max proc file

[PATCH] RFC - add prealloc_max proc file

[Jaroslav Kysela]

Hi,

	I would like to propose to add prealloc_max proc file to give a 
hint, what's the maximum allowed DMA buffer size in bytes:

diff -r 1551b696524c core/pcm_memory.c
--- a/core/pcm_memory.c	Fri Oct 06 09:34:20 2006 +0200
+++ b/core/pcm_memory.c	Fri Oct 06 10:26:15 2006 +0200
@@ -101,6 +101,8 @@ int snd_pcm_lib_preallocate_free(struct 
 {
 	snd_pcm_lib_preallocate_dma_free(substream);
 #ifdef CONFIG_SND_VERBOSE_PROCFS
+	snd_info_free_entry(substream->proc_prealloc_max_entry);
+	substream->proc_prealloc_max_entry = NULL;
 	snd_info_free_entry(substream->proc_prealloc_entry);
 	substream->proc_prealloc_entry = NULL;
 #endif
@@ -139,6 +141,18 @@ static void snd_pcm_lib_preallocate_proc
 {
 	struct snd_pcm_substream *substream = entry->private_data;
 	snd_iprintf(buffer, "%lu\n", (unsigned long) substream->dma_buffer.bytes / 1024);
+}
+
+/*
+ * read callback for prealloc_max proc file
+ *
+ * prints the maximum allowed size in kB.
+ */
+static void snd_pcm_lib_preallocate_max_proc_read(struct snd_info_entry *entry,
+						  struct snd_info_buffer *buffer)
+{
+	struct snd_pcm_substream *substream = entry->private_data;
+	snd_iprintf(buffer, "%lu\n", (unsigned long) substream->dma_max / 1024);
 }
 
 /*
@@ -203,6 +217,15 @@ static inline void preallocate_info_init
 		}
 	}
 	substream->proc_prealloc_entry = entry;
+	if ((entry = snd_info_create_card_entry(substream->pcm->card, "prealloc_max", substream->proc_root)) != NULL) {
+		entry->c.text.read = snd_pcm_lib_preallocate_max_proc_read;
+		entry->private_data = substream;
+		if (snd_info_register(entry) < 0) {
+			snd_info_free_entry(entry);
+			entry = NULL;
+		}
+	}
+	substream->proc_prealloc_max_entry = entry;
 }
 
 #else /* !CONFIG_SND_VERBOSE_PROCFS */
diff -r 1551b696524c include/pcm.h
--- a/include/pcm.h	Fri Oct 06 09:34:20 2006 +0200
+++ b/include/pcm.h	Fri Oct 06 10:26:15 2006 +0200
@@ -383,6 +383,7 @@ struct snd_pcm_substream {
 	struct snd_info_entry *proc_sw_params_entry;
 	struct snd_info_entry *proc_status_entry;
 	struct snd_info_entry *proc_prealloc_entry;
+	struct snd_info_entry *proc_prealloc_max_entry;
 #endif
 	/* misc flags */
 	unsigned int hw_opened: 1;

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SUSE Labs

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [PATCH] RFC - add prealloc_max proc file

[Clemens Ladisch]

Jaroslav Kysela wrote:
> I would like to propose to add prealloc_max proc file to give a 
> hint, what's the maximum allowed DMA buffer size in bytes:

You mean, users are not supposed to do a binary search with
the prealloc file?  ;-)

> ...

Looks good.


BTW: Shouldn't all(?) these proc files eventually go to sysfs?


Regards,
Clemens

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [PATCH] RFC - add prealloc_max proc file

[Takashi Iwai]

At Fri, 06 Oct 2006 11:15:04 +0200,
Clemens Ladisch wrote:
> 
> Jaroslav Kysela wrote:
> > I would like to propose to add prealloc_max proc file to give a 
> > hint, what's the maximum allowed DMA buffer size in bytes:
> 
> You mean, users are not supposed to do a binary search with
> the prealloc file?  ;-)
> 
> > ...
> 
> Looks good.
> 
> 
> BTW: Shouldn't all(?) these proc files eventually go to sysfs?

There is already a function to add a sysfs file
(snd_add_device_sysfs_file()), so it must be pretty easy.


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [PATCH] RFC - add prealloc_max proc file

[Jaroslav Kysela]

On Fri, 6 Oct 2006, Takashi Iwai wrote:

> At Fri, 06 Oct 2006 11:15:04 +0200,
> Clemens Ladisch wrote:
> > 
> > Jaroslav Kysela wrote:
> > > I would like to propose to add prealloc_max proc file to give a 
> > > hint, what's the maximum allowed DMA buffer size in bytes:
> > 
> > You mean, users are not supposed to do a binary search with
> > the prealloc file?  ;-)
> > 
> > > ...
> > 
> > Looks good.
> > 
> > 
> > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> 
> There is already a function to add a sysfs file
> (snd_add_device_sysfs_file()), so it must be pretty easy.

Yes and no. An layer for concrete device/substream should be added.
Anyway, actually we can move most of values from proc to sysfs for 2.6.
Any problems with it?

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SUSE Labs

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [PATCH] RFC - add prealloc_max proc file

[Takashi Iwai]

At Fri, 6 Oct 2006 13:36:51 +0200 (CEST),
Jaroslav Kysela wrote:
> 
> On Fri, 6 Oct 2006, Takashi Iwai wrote:
> 
> > At Fri, 06 Oct 2006 11:15:04 +0200,
> > Clemens Ladisch wrote:
> > > 
> > > Jaroslav Kysela wrote:
> > > > I would like to propose to add prealloc_max proc file to give a 
> > > > hint, what's the maximum allowed DMA buffer size in bytes:
> > > 
> > > You mean, users are not supposed to do a binary search with
> > > the prealloc file?  ;-)
> > > 
> > > > ...
> > > 
> > > Looks good.
> > > 
> > > 
> > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > 
> > There is already a function to add a sysfs file
> > (snd_add_device_sysfs_file()), so it must be pretty easy.
> 
> Yes and no. An layer for concrete device/substream should be added.

Right, it's missing.

> Anyway, actually we can move most of values from proc to sysfs for 2.6.
> Any problems with it?

Proc files for showing information are OK, but controlling the driver
over proc file isn't a recommended way on 2.6 kernel.  That's what
sysfs is for.

Anyway it's no big deal but just TODO.


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Jaroslav Kysela]

On Fri, 6 Oct 2006, Takashi Iwai wrote:

> > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > 
> > > There is already a function to add a sysfs file
> > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > 
> > Yes and no. An layer for concrete device/substream should be added.
> 
> Right, it's missing.
> 
> > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > Any problems with it?
> 
> Proc files for showing information are OK, but controlling the driver
> over proc file isn't a recommended way on 2.6 kernel.  That's what
> sysfs is for.
> 
> Anyway it's no big deal but just TODO.

I'd like to create similar layer as in proc, but it seems that we must 
place cards root somewhere. Do you think that /sys/class/soundcard is good 
place and name for this root?

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SUSE Labs

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Takashi Iwai]

At Fri, 6 Oct 2006 15:19:58 +0200 (CEST),
Jaroslav Kysela wrote:
> 
> On Fri, 6 Oct 2006, Takashi Iwai wrote:
> 
> > > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > > 
> > > > There is already a function to add a sysfs file
> > > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > > 
> > > Yes and no. An layer for concrete device/substream should be added.
> > 
> > Right, it's missing.
> > 
> > > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > > Any problems with it?
> > 
> > Proc files for showing information are OK, but controlling the driver
> > over proc file isn't a recommended way on 2.6 kernel.  That's what
> > sysfs is for.
> > 
> > Anyway it's no big deal but just TODO.
> 
> I'd like to create similar layer as in proc, but it seems that we must 
> place cards root somewhere. Do you think that /sys/class/soundcard is good 
> place and name for this root?

I don't think so.  /sys/class are referred by HAL to probe the device
files.  Adding a root would simply confuse the system.


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Jaroslav Kysela]

On Fri, 6 Oct 2006, Takashi Iwai wrote:

> At Fri, 6 Oct 2006 15:19:58 +0200 (CEST),
> Jaroslav Kysela wrote:
> > 
> > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > 
> > > > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > > > 
> > > > > There is already a function to add a sysfs file
> > > > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > > > 
> > > > Yes and no. An layer for concrete device/substream should be added.
> > > 
> > > Right, it's missing.
> > > 
> > > > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > > > Any problems with it?
> > > 
> > > Proc files for showing information are OK, but controlling the driver
> > > over proc file isn't a recommended way on 2.6 kernel.  That's what
> > > sysfs is for.
> > > 
> > > Anyway it's no big deal but just TODO.
> > 
> > I'd like to create similar layer as in proc, but it seems that we must 
> > place cards root somewhere. Do you think that /sys/class/soundcard is good 
> > place and name for this root?
> 
> I don't think so.  /sys/class are referred by HAL to probe the device
> files.  Adding a root would simply confuse the system.

Hmm. Any suggestion? /sys/bus or /sys/devices does not seem also as a 
right place.

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SUSE Labs

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Takashi Iwai]

At Fri, 6 Oct 2006 15:34:30 +0200 (CEST),
Jaroslav Kysela wrote:
> 
> On Fri, 6 Oct 2006, Takashi Iwai wrote:
> 
> > At Fri, 6 Oct 2006 15:19:58 +0200 (CEST),
> > Jaroslav Kysela wrote:
> > > 
> > > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > > 
> > > > > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > > > > 
> > > > > > There is already a function to add a sysfs file
> > > > > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > > > > 
> > > > > Yes and no. An layer for concrete device/substream should be added.
> > > > 
> > > > Right, it's missing.
> > > > 
> > > > > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > > > > Any problems with it?
> > > > 
> > > > Proc files for showing information are OK, but controlling the driver
> > > > over proc file isn't a recommended way on 2.6 kernel.  That's what
> > > > sysfs is for.
> > > > 
> > > > Anyway it's no big deal but just TODO.
> > > 
> > > I'd like to create similar layer as in proc, but it seems that we must 
> > > place cards root somewhere. Do you think that /sys/class/soundcard is good 
> > > place and name for this root?
> > 
> > I don't think so.  /sys/class are referred by HAL to probe the device
> > files.  Adding a root would simply confuse the system.
> 
> Hmm. Any suggestion? /sys/bus or /sys/devices does not seem also as a 
> right place.

We can create another class for the sound-root.  But, better to ask
sysfs guys before starting working on it.

Alternatively, we can put the tree under controlC* entry since control
is corresponding to each card in ALSA.  Actually, snd_card instance is
_the_ container of control elements.


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Jaroslav Kysela]

On Fri, 6 Oct 2006, Takashi Iwai wrote:

> At Fri, 6 Oct 2006 15:34:30 +0200 (CEST),
> Jaroslav Kysela wrote:
> > 
> > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > 
> > > At Fri, 6 Oct 2006 15:19:58 +0200 (CEST),
> > > Jaroslav Kysela wrote:
> > > > 
> > > > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > > > 
> > > > > > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > > > > > 
> > > > > > > There is already a function to add a sysfs file
> > > > > > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > > > > > 
> > > > > > Yes and no. An layer for concrete device/substream should be added.
> > > > > 
> > > > > Right, it's missing.
> > > > > 
> > > > > > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > > > > > Any problems with it?
> > > > > 
> > > > > Proc files for showing information are OK, but controlling the driver
> > > > > over proc file isn't a recommended way on 2.6 kernel.  That's what
> > > > > sysfs is for.
> > > > > 
> > > > > Anyway it's no big deal but just TODO.
> > > > 
> > > > I'd like to create similar layer as in proc, but it seems that we must 
> > > > place cards root somewhere. Do you think that /sys/class/soundcard is good 
> > > > place and name for this root?
> > > 
> > > I don't think so.  /sys/class are referred by HAL to probe the device
> > > files.  Adding a root would simply confuse the system.
> > 
> > Hmm. Any suggestion? /sys/bus or /sys/devices does not seem also as a 
> > right place.
> 
> We can create another class for the sound-root.  But, better to ask
> sysfs guys before starting working on it.
> 
> Alternatively, we can put the tree under controlC* entry since control
> is corresponding to each card in ALSA.  Actually, snd_card instance is
> _the_ container of control elements.

Looking to the udev code, creating /sys/class/soundcard might make sense.
It seems that they're handling /sys/class/net also differently than other 
standard /dev devices.

The main purpose of this class should be renaming card id.

Anyway, let's wait if someone replies on lkml.

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SUSE Labs

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: SYSFS, was Re: [PATCH] RFC - add prealloc_max proc file

[Takashi Iwai]

At Fri, 6 Oct 2006 19:30:32 +0200 (CEST),
Jaroslav Kysela wrote:
> 
> On Fri, 6 Oct 2006, Takashi Iwai wrote:
> 
> > At Fri, 6 Oct 2006 15:34:30 +0200 (CEST),
> > Jaroslav Kysela wrote:
> > > 
> > > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > > 
> > > > At Fri, 6 Oct 2006 15:19:58 +0200 (CEST),
> > > > Jaroslav Kysela wrote:
> > > > > 
> > > > > On Fri, 6 Oct 2006, Takashi Iwai wrote:
> > > > > 
> > > > > > > > > BTW: Shouldn't all(?) these proc files eventually go to sysfs?
> > > > > > > > 
> > > > > > > > There is already a function to add a sysfs file
> > > > > > > > (snd_add_device_sysfs_file()), so it must be pretty easy.
> > > > > > > 
> > > > > > > Yes and no. An layer for concrete device/substream should be added.
> > > > > > 
> > > > > > Right, it's missing.
> > > > > > 
> > > > > > > Anyway, actually we can move most of values from proc to sysfs for 2.6.
> > > > > > > Any problems with it?
> > > > > > 
> > > > > > Proc files for showing information are OK, but controlling the driver
> > > > > > over proc file isn't a recommended way on 2.6 kernel.  That's what
> > > > > > sysfs is for.
> > > > > > 
> > > > > > Anyway it's no big deal but just TODO.
> > > > > 
> > > > > I'd like to create similar layer as in proc, but it seems that we must 
> > > > > place cards root somewhere. Do you think that /sys/class/soundcard is good 
> > > > > place and name for this root?
> > > > 
> > > > I don't think so.  /sys/class are referred by HAL to probe the device
> > > > files.  Adding a root would simply confuse the system.
> > > 
> > > Hmm. Any suggestion? /sys/bus or /sys/devices does not seem also as a 
> > > right place.
> > 
> > We can create another class for the sound-root.  But, better to ask
> > sysfs guys before starting working on it.
> > 
> > Alternatively, we can put the tree under controlC* entry since control
> > is corresponding to each card in ALSA.  Actually, snd_card instance is
> > _the_ container of control elements.
> 
> Looking to the udev code, creating /sys/class/soundcard might make sense.
> It seems that they're handling /sys/class/net also differently than other 
> standard /dev devices.

It's not only udev but also HAL that look for /sys...


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV