All of lore.kernel.org
 help / color / mirror / Atom feed
* snd-ctxfi
@ 2009-05-30  3:22 The Source
  2009-05-30  7:00 ` snd-ctxfi Takashi Iwai
  0 siblings, 1 reply; 8+ messages in thread
From: The Source @ 2009-05-30  3:22 UTC (permalink / raw)
  To: alsa-devel

Hello. I tested the snd-ctxfi driver and found no problems (except no 
support for 5.1 and external console, but original ctxfi does not 
support them either). A minor problem: kernel oops reporter periodically 
pops up with message about using a timer in a wrong place. I'll post 
here this message when I get it next time.

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-05-30  3:22 snd-ctxfi The Source
@ 2009-05-30  7:00 ` Takashi Iwai
  2009-05-30 16:22   ` snd-ctxfi The Source
  0 siblings, 1 reply; 8+ messages in thread
From: Takashi Iwai @ 2009-05-30  7:00 UTC (permalink / raw)
  To: The Source; +Cc: alsa-devel

At Sat, 30 May 2009 07:22:43 +0400,
The Source wrote:
> 
> Hello. I tested the snd-ctxfi driver and found no problems (except no 
> support for 5.1 and external console, but original ctxfi does not 
> support them either). A minor problem: kernel oops reporter periodically 
> pops up with message about using a timer in a wrong place.

Oh, that's bad.

> I'll post 
> here this message when I get it next time.

Yes, please, that'll be helpful.


thanks,

Takashi

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-05-30  7:00 ` snd-ctxfi Takashi Iwai
@ 2009-05-30 16:22   ` The Source
  2009-06-01  9:10     ` snd-ctxfi Takashi Iwai
  0 siblings, 1 reply; 8+ messages in thread
From: The Source @ 2009-05-30 16:22 UTC (permalink / raw)
  To: Takashi Iwai; +Cc: alsa-devel

On 30.05.2009 11:00, Takashi Iwai wrote:
> At Sat, 30 May 2009 07:22:43 +0400,
> The Source wrote:
>    
>> Hello. I tested the snd-ctxfi driver and found no problems (except no
>> support for 5.1 and external console, but original ctxfi does not
>> support them either). A minor problem: kernel oops reporter periodically
>> pops up with message about using a timer in a wrong place.
>>      
> Oh, that's bad.
>
>    
>> I'll post
>> here this message when I get it next time.
>>      
> Yes, please, that'll be helpful.
>
>
> thanks,
>
> Takashi
>
>    
Here's the message:

BUG: sleeping function called from invalid context at mm/slub.c:1599
in_atomic(): 0, irqs_disabled(): 1, pid: 32065, name: xine
Pid: 32065, comm: xine Tainted: P           2.6.29.4-75.fc10.x86_64 #1
Call Trace:
  [<ffffffff81040685>] __might_sleep+0x105/0x10a
  [<ffffffff810c9fae>] kmem_cache_alloc+0x32/0xe2
  [<ffffffffa08e3110>] ct_vm_map+0xfa/0x19e [snd_ctxfi]
  [<ffffffffa08e1a07>] ct_map_audio_buffer+0x4c/0x76 [snd_ctxfi]
  [<ffffffffa08e2aa5>] atc_pcm_playback_prepare+0x1d7/0x2a8 [snd_ctxfi]
  [<ffffffff8105ef3f>] ? up_read+0x9/0xb
  [<ffffffff81186b61>] ? __up_read+0x7c/0x87
  [<ffffffffa08e36a6>] ct_pcm_playback_prepare+0x39/0x60 [snd_ctxfi]
  [<ffffffffa0886bcb>] snd_pcm_do_prepare+0x16/0x28 [snd_pcm]
  [<ffffffffa08867c7>] snd_pcm_action_single+0x2d/0x5b [snd_pcm]
  [<ffffffffa08881f3>] snd_pcm_action_nonatomic+0x52/0x6a [snd_pcm]
  [<ffffffffa088a723>] snd_pcm_common_ioctl1+0x404/0xc79 [snd_pcm]
  [<ffffffff810c52c8>] ? alloc_pages_current+0xb9/0xc2
  [<ffffffff810c9402>] ? new_slab+0x1a5/0x1cb
  [<ffffffff810ab9ea>] ? vma_prio_tree_insert+0x23/0xc1
  [<ffffffffa088b411>] snd_pcm_playback_ioctl1+0x213/0x230 [snd_pcm]
  [<ffffffff810b6c20>] ? mmap_region+0x397/0x4c9
  [<ffffffffa088bd9b>] snd_pcm_playback_ioctl+0x2e/0x36 [snd_pcm]
  [<ffffffff810ddc64>] vfs_ioctl+0x2a/0x78
  [<ffffffff810de130>] do_vfs_ioctl+0x462/0x4a2
  [<ffffffff81029cef>] ? default_spin_lock_flags+0x9/0xe
  [<ffffffff81374647>] ? trace_hardirqs_off_thunk+0x3a/0x6c
  [<ffffffff810de1c5>] sys_ioctl+0x55/0x77
  [<ffffffff8101133a>] system_call_fastpath+0x16/0x1b

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-05-30 16:22   ` snd-ctxfi The Source
@ 2009-06-01  9:10     ` Takashi Iwai
  2009-06-02  1:47       ` snd-ctxfi Lee Revell
  0 siblings, 1 reply; 8+ messages in thread
From: Takashi Iwai @ 2009-06-01  9:10 UTC (permalink / raw)
  To: The Source; +Cc: alsa-devel

At Sat, 30 May 2009 20:22:46 +0400,
The Source wrote:
> 
> On 30.05.2009 11:00, Takashi Iwai wrote:
> > At Sat, 30 May 2009 07:22:43 +0400,
> > The Source wrote:
> >    
> >> Hello. I tested the snd-ctxfi driver and found no problems (except no
> >> support for 5.1 and external console, but original ctxfi does not
> >> support them either). A minor problem: kernel oops reporter periodically
> >> pops up with message about using a timer in a wrong place.
> >>      
> > Oh, that's bad.
> >
> >    
> >> I'll post
> >> here this message when I get it next time.
> >>      
> > Yes, please, that'll be helpful.
> >
> >
> > thanks,
> >
> > Takashi
> >
> >    
> Here's the message:
> 
> BUG: sleeping function called from invalid context at mm/slub.c:1599
> in_atomic(): 0, irqs_disabled(): 1, pid: 32065, name: xine
> Pid: 32065, comm: xine Tainted: P           2.6.29.4-75.fc10.x86_64 #1
> Call Trace:
>   [<ffffffff81040685>] __might_sleep+0x105/0x10a
>   [<ffffffff810c9fae>] kmem_cache_alloc+0x32/0xe2
>   [<ffffffffa08e3110>] ct_vm_map+0xfa/0x19e [snd_ctxfi]
>   [<ffffffffa08e1a07>] ct_map_audio_buffer+0x4c/0x76 [snd_ctxfi]
>   [<ffffffffa08e2aa5>] atc_pcm_playback_prepare+0x1d7/0x2a8 [snd_ctxfi]
>   [<ffffffff8105ef3f>] ? up_read+0x9/0xb
>   [<ffffffff81186b61>] ? __up_read+0x7c/0x87
>   [<ffffffffa08e36a6>] ct_pcm_playback_prepare+0x39/0x60 [snd_ctxfi]
>   [<ffffffffa0886bcb>] snd_pcm_do_prepare+0x16/0x28 [snd_pcm]
>   [<ffffffffa08867c7>] snd_pcm_action_single+0x2d/0x5b [snd_pcm]
>   [<ffffffffa08881f3>] snd_pcm_action_nonatomic+0x52/0x6a [snd_pcm]
>   [<ffffffffa088a723>] snd_pcm_common_ioctl1+0x404/0xc79 [snd_pcm]
>   [<ffffffff810c52c8>] ? alloc_pages_current+0xb9/0xc2
>   [<ffffffff810c9402>] ? new_slab+0x1a5/0x1cb
>   [<ffffffff810ab9ea>] ? vma_prio_tree_insert+0x23/0xc1
>   [<ffffffffa088b411>] snd_pcm_playback_ioctl1+0x213/0x230 [snd_pcm]
>   [<ffffffff810b6c20>] ? mmap_region+0x397/0x4c9
>   [<ffffffffa088bd9b>] snd_pcm_playback_ioctl+0x2e/0x36 [snd_pcm]
>   [<ffffffff810ddc64>] vfs_ioctl+0x2a/0x78
>   [<ffffffff810de130>] do_vfs_ioctl+0x462/0x4a2
>   [<ffffffff81029cef>] ? default_spin_lock_flags+0x9/0xe
>   [<ffffffff81374647>] ? trace_hardirqs_off_thunk+0x3a/0x6c
>   [<ffffffff810de1c5>] sys_ioctl+0x55/0x77
>   [<ffffffff8101133a>] system_call_fastpath+0x16/0x1b

Thanks.  So, something wrong in the mmap handling.
I'll take a look later in this week.


Takashi

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-06-01  9:10     ` snd-ctxfi Takashi Iwai
@ 2009-06-02  1:47       ` Lee Revell
  2009-06-02  6:09         ` snd-ctxfi Takashi Iwai
  0 siblings, 1 reply; 8+ messages in thread
From: Lee Revell @ 2009-06-02  1:47 UTC (permalink / raw)
  To: Takashi Iwai; +Cc: alsa-devel, The Source

On Mon, Jun 1, 2009 at 5:10 AM, Takashi Iwai <tiwai@suse.de> wrote:
> Thanks.  So, something wrong in the mmap handling.
> I'll take a look later in this week.

Looks like ct_map_audio_buffer calls ct_vm_map under spinlock;
ct_vm_map calls get_vm_block which calls kzalloc with GFP_KERNEL.

Lee

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-06-02  1:47       ` snd-ctxfi Lee Revell
@ 2009-06-02  6:09         ` Takashi Iwai
  2009-06-04  3:48           ` snd-ctxfi The Source
  0 siblings, 1 reply; 8+ messages in thread
From: Takashi Iwai @ 2009-06-02  6:09 UTC (permalink / raw)
  To: Lee Revell; +Cc: alsa-devel, The Source

At Mon, 1 Jun 2009 21:47:02 -0400,
Lee Revell wrote:
> 
> On Mon, Jun 1, 2009 at 5:10 AM, Takashi Iwai <tiwai@suse.de> wrote:
> > Thanks.  So, something wrong in the mmap handling.
> > I'll take a look later in this week.
> 
> Looks like ct_map_audio_buffer calls ct_vm_map under spinlock;
> ct_vm_map calls get_vm_block which calls kzalloc with GFP_KERNEL.

Indeed.  Does the patch below fix the problem?


thanks,

Takashi

---
diff --git a/sound/pci/ctxfi/ctatc.c b/sound/pci/ctxfi/ctatc.c
index ead104e..1a4bb35 100644
--- a/sound/pci/ctxfi/ctatc.c
+++ b/sound/pci/ctxfi/ctatc.c
@@ -119,7 +119,6 @@ atc_pcm_release_resources(struct ct_atc *atc, struct ct_atc_pcm *apcm);
 
 static int ct_map_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm *apcm)
 {
-	unsigned long flags;
 	struct snd_pcm_runtime *runtime;
 	struct ct_vm *vm;
 
@@ -129,9 +128,7 @@ static int ct_map_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm *apcm)
 	runtime = apcm->substream->runtime;
 	vm = atc->vm;
 
-	spin_lock_irqsave(&atc->vm_lock, flags);
 	apcm->vm_block = vm->map(vm, runtime->dma_area, runtime->dma_bytes);
-	spin_unlock_irqrestore(&atc->vm_lock, flags);
 
 	if (NULL == apcm->vm_block)
 		return -ENOENT;
@@ -141,7 +138,6 @@ static int ct_map_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm *apcm)
 
 static void ct_unmap_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm *apcm)
 {
-	unsigned long flags;
 	struct ct_vm *vm;
 
 	if (NULL == apcm->vm_block)
@@ -149,9 +145,7 @@ static void ct_unmap_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm *apcm)
 
 	vm = atc->vm;
 
-	spin_lock_irqsave(&atc->vm_lock, flags);
 	vm->unmap(vm, apcm->vm_block);
-	spin_unlock_irqrestore(&atc->vm_lock, flags);
 
 	apcm->vm_block = NULL;
 }
@@ -161,9 +155,7 @@ static unsigned long atc_get_ptp_phys(struct ct_atc *atc, int index)
 	struct ct_vm *vm;
 	void *kvirt_addr;
 	unsigned long phys_addr;
-	unsigned long flags;
 
-	spin_lock_irqsave(&atc->vm_lock, flags);
 	vm = atc->vm;
 	kvirt_addr = vm->get_ptp_virt(vm, index);
 	if (kvirt_addr == NULL)
@@ -171,8 +163,6 @@ static unsigned long atc_get_ptp_phys(struct ct_atc *atc, int index)
 	else
 		phys_addr = virt_to_phys(kvirt_addr);
 
-	spin_unlock_irqrestore(&atc->vm_lock, flags);
-
 	return phys_addr;
 }
 
@@ -1562,7 +1552,6 @@ int ct_atc_create(struct snd_card *card, struct pci_dev *pci,
 	atc_set_ops(atc);
 
 	spin_lock_init(&atc->atc_lock);
-	spin_lock_init(&atc->vm_lock);
 
 	/* Find card model */
 	err = atc_identify_card(atc);
diff --git a/sound/pci/ctxfi/ctatc.h b/sound/pci/ctxfi/ctatc.h
index 286c993..a7b0ec2 100644
--- a/sound/pci/ctxfi/ctatc.h
+++ b/sound/pci/ctxfi/ctatc.h
@@ -101,7 +101,6 @@ struct ct_atc {
 	unsigned long (*get_ptp_phys)(struct ct_atc *atc, int index);
 
 	spinlock_t atc_lock;
-	spinlock_t vm_lock;
 
 	int (*pcm_playback_prepare)(struct ct_atc *atc,
 				    struct ct_atc_pcm *apcm);
diff --git a/sound/pci/ctxfi/ctvmem.c b/sound/pci/ctxfi/ctvmem.c
index cecf77e..363b67e 100644
--- a/sound/pci/ctxfi/ctvmem.c
+++ b/sound/pci/ctxfi/ctvmem.c
@@ -35,25 +35,27 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
 	struct ct_vm_block *block = NULL, *entry = NULL;
 	struct list_head *pos = NULL;
 
+	mutex_lock(&vm->lock);
 	list_for_each(pos, &vm->unused) {
 		entry = list_entry(pos, struct ct_vm_block, list);
 		if (entry->size >= size)
 			break; /* found a block that is big enough */
 	}
 	if (pos == &vm->unused)
-		return NULL;
+		goto out;
 
 	if (entry->size == size) {
 		/* Move the vm node from unused list to used list directly */
 		list_del(&entry->list);
 		list_add(&entry->list, &vm->used);
 		vm->size -= size;
-		return entry;
+		block = entry;
+		goto out;
 	}
 
 	block = kzalloc(sizeof(*block), GFP_KERNEL);
 	if (NULL == block)
-		return NULL;
+		goto out;
 
 	block->addr = entry->addr;
 	block->size = size;
@@ -62,6 +64,8 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
 	entry->size -= size;
 	vm->size -= size;
 
+ out:
+	mutex_unlock(&vm->lock);
 	return block;
 }
 
@@ -70,6 +74,7 @@ static void put_vm_block(struct ct_vm *vm, struct ct_vm_block *block)
 	struct ct_vm_block *entry = NULL, *pre_ent = NULL;
 	struct list_head *pos = NULL, *pre = NULL;
 
+	mutex_lock(&vm->lock);
 	list_del(&block->list);
 	vm->size += block->size;
 
@@ -106,6 +111,7 @@ static void put_vm_block(struct ct_vm *vm, struct ct_vm_block *block)
 		pos = pre;
 		pre = pos->prev;
 	}
+	mutex_unlock(&vm->lock);
 }
 
 /* Map host addr (kmalloced/vmalloced) to device logical addr. */
@@ -191,6 +197,8 @@ int ct_vm_create(struct ct_vm **rvm)
 	if (NULL == vm)
 		return -ENOMEM;
 
+	mutex_init(&vm->lock);
+
 	/* Allocate page table pages */
 	for (i = 0; i < CT_PTP_NUM; i++) {
 		vm->ptp[i] = kmalloc(PAGE_SIZE, GFP_KERNEL);
diff --git a/sound/pci/ctxfi/ctvmem.h b/sound/pci/ctxfi/ctvmem.h
index 4eb5bdd..618952e 100644
--- a/sound/pci/ctxfi/ctvmem.h
+++ b/sound/pci/ctxfi/ctvmem.h
@@ -20,7 +20,7 @@
 
 #define CT_PTP_NUM	1	/* num of device page table pages */
 
-#include <linux/spinlock.h>
+#include <linux/mutex.h>
 #include <linux/list.h>
 
 struct ct_vm_block {
@@ -35,6 +35,7 @@ struct ct_vm {
 	unsigned int size;		/* Available addr space in bytes */
 	struct list_head unused;	/* List of unused blocks */
 	struct list_head used;		/* List of used blocks */
+	struct mutex lock;
 
 	/* Map host addr (kmalloced/vmalloced) to device logical addr. */
 	struct ct_vm_block *(*map)(struct ct_vm *, void *host_addr, int size);

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-06-02  6:09         ` snd-ctxfi Takashi Iwai
@ 2009-06-04  3:48           ` The Source
  2009-06-04  5:29             ` snd-ctxfi Takashi Iwai
  0 siblings, 1 reply; 8+ messages in thread
From: The Source @ 2009-06-04  3:48 UTC (permalink / raw)
  To: Takashi Iwai; +Cc: alsa-devel

2009/6/2 Takashi Iwai <tiwai@suse.de>

> At Mon, 1 Jun 2009 21:47:02 -0400,
> Lee Revell wrote:
> >
> > On Mon, Jun 1, 2009 at 5:10 AM, Takashi Iwai <tiwai@suse.de> wrote:
> > > Thanks.  So, something wrong in the mmap handling.
> > > I'll take a look later in this week.
> >
> > Looks like ct_map_audio_buffer calls ct_vm_map under spinlock;
> > ct_vm_map calls get_vm_block which calls kzalloc with GFP_KERNEL.
>
> Indeed.  Does the patch below fix the problem?
>
>
> thanks,
>
> Takashi
>
> ---
> diff --git a/sound/pci/ctxfi/ctatc.c b/sound/pci/ctxfi/ctatc.c
> index ead104e..1a4bb35 100644
> --- a/sound/pci/ctxfi/ctatc.c
> +++ b/sound/pci/ctxfi/ctatc.c
> @@ -119,7 +119,6 @@ atc_pcm_release_resources(struct ct_atc *atc, struct
> ct_atc_pcm *apcm);
>
>  static int ct_map_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm
> *apcm)
>  {
> -       unsigned long flags;
>        struct snd_pcm_runtime *runtime;
>        struct ct_vm *vm;
>
> @@ -129,9 +128,7 @@ static int ct_map_audio_buffer(struct ct_atc *atc,
> struct ct_atc_pcm *apcm)
>        runtime = apcm->substream->runtime;
>        vm = atc->vm;
>
> -       spin_lock_irqsave(&atc->vm_lock, flags);
>        apcm->vm_block = vm->map(vm, runtime->dma_area, runtime->dma_bytes);
> -       spin_unlock_irqrestore(&atc->vm_lock, flags);
>
>        if (NULL == apcm->vm_block)
>                return -ENOENT;
> @@ -141,7 +138,6 @@ static int ct_map_audio_buffer(struct ct_atc *atc,
> struct ct_atc_pcm *apcm)
>
>  static void ct_unmap_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm
> *apcm)
>  {
> -       unsigned long flags;
>        struct ct_vm *vm;
>
>        if (NULL == apcm->vm_block)
> @@ -149,9 +145,7 @@ static void ct_unmap_audio_buffer(struct ct_atc *atc,
> struct ct_atc_pcm *apcm)
>
>        vm = atc->vm;
>
> -       spin_lock_irqsave(&atc->vm_lock, flags);
>        vm->unmap(vm, apcm->vm_block);
> -       spin_unlock_irqrestore(&atc->vm_lock, flags);
>
>        apcm->vm_block = NULL;
>  }
> @@ -161,9 +155,7 @@ static unsigned long atc_get_ptp_phys(struct ct_atc
> *atc, int index)
>        struct ct_vm *vm;
>        void *kvirt_addr;
>        unsigned long phys_addr;
> -       unsigned long flags;
>
> -       spin_lock_irqsave(&atc->vm_lock, flags);
>        vm = atc->vm;
>        kvirt_addr = vm->get_ptp_virt(vm, index);
>        if (kvirt_addr == NULL)
> @@ -171,8 +163,6 @@ static unsigned long atc_get_ptp_phys(struct ct_atc
> *atc, int index)
>        else
>                phys_addr = virt_to_phys(kvirt_addr);
>
> -       spin_unlock_irqrestore(&atc->vm_lock, flags);
> -
>        return phys_addr;
>  }
>
> @@ -1562,7 +1552,6 @@ int ct_atc_create(struct snd_card *card, struct
> pci_dev *pci,
>        atc_set_ops(atc);
>
>        spin_lock_init(&atc->atc_lock);
> -       spin_lock_init(&atc->vm_lock);
>
>        /* Find card model */
>        err = atc_identify_card(atc);
> diff --git a/sound/pci/ctxfi/ctatc.h b/sound/pci/ctxfi/ctatc.h
> index 286c993..a7b0ec2 100644
> --- a/sound/pci/ctxfi/ctatc.h
> +++ b/sound/pci/ctxfi/ctatc.h
> @@ -101,7 +101,6 @@ struct ct_atc {
>        unsigned long (*get_ptp_phys)(struct ct_atc *atc, int index);
>
>        spinlock_t atc_lock;
> -       spinlock_t vm_lock;
>
>        int (*pcm_playback_prepare)(struct ct_atc *atc,
>                                    struct ct_atc_pcm *apcm);
> diff --git a/sound/pci/ctxfi/ctvmem.c b/sound/pci/ctxfi/ctvmem.c
> index cecf77e..363b67e 100644
> --- a/sound/pci/ctxfi/ctvmem.c
> +++ b/sound/pci/ctxfi/ctvmem.c
> @@ -35,25 +35,27 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
>        struct ct_vm_block *block = NULL, *entry = NULL;
>        struct list_head *pos = NULL;
>
> +       mutex_lock(&vm->lock);
>        list_for_each(pos, &vm->unused) {
>                entry = list_entry(pos, struct ct_vm_block, list);
>                if (entry->size >= size)
>                        break; /* found a block that is big enough */
>        }
>        if (pos == &vm->unused)
> -               return NULL;
> +               goto out;
>
>        if (entry->size == size) {
>                /* Move the vm node from unused list to used list directly
> */
>                list_del(&entry->list);
>                list_add(&entry->list, &vm->used);
>                vm->size -= size;
> -               return entry;
> +               block = entry;
> +               goto out;
>        }
>
>        block = kzalloc(sizeof(*block), GFP_KERNEL);
>        if (NULL == block)
> -               return NULL;
> +               goto out;
>
>        block->addr = entry->addr;
>        block->size = size;
> @@ -62,6 +64,8 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
>        entry->size -= size;
>        vm->size -= size;
>
> + out:
> +       mutex_unlock(&vm->lock);
>        return block;
>  }
>
> @@ -70,6 +74,7 @@ static void put_vm_block(struct ct_vm *vm, struct
> ct_vm_block *block)
>        struct ct_vm_block *entry = NULL, *pre_ent = NULL;
>        struct list_head *pos = NULL, *pre = NULL;
>
> +       mutex_lock(&vm->lock);
>        list_del(&block->list);
>        vm->size += block->size;
>
> @@ -106,6 +111,7 @@ static void put_vm_block(struct ct_vm *vm, struct
> ct_vm_block *block)
>                pos = pre;
>                pre = pos->prev;
>        }
> +       mutex_unlock(&vm->lock);
>  }
>
>  /* Map host addr (kmalloced/vmalloced) to device logical addr. */
> @@ -191,6 +197,8 @@ int ct_vm_create(struct ct_vm **rvm)
>        if (NULL == vm)
>                return -ENOMEM;
>
> +       mutex_init(&vm->lock);
> +
>        /* Allocate page table pages */
>        for (i = 0; i < CT_PTP_NUM; i++) {
>                vm->ptp[i] = kmalloc(PAGE_SIZE, GFP_KERNEL);
> diff --git a/sound/pci/ctxfi/ctvmem.h b/sound/pci/ctxfi/ctvmem.h
> index 4eb5bdd..618952e 100644
> --- a/sound/pci/ctxfi/ctvmem.h
> +++ b/sound/pci/ctxfi/ctvmem.h
> @@ -20,7 +20,7 @@
>
>  #define CT_PTP_NUM     1       /* num of device page table pages */
>
> -#include <linux/spinlock.h>
> +#include <linux/mutex.h>
>  #include <linux/list.h>
>
>  struct ct_vm_block {
> @@ -35,6 +35,7 @@ struct ct_vm {
>        unsigned int size;              /* Available addr space in bytes */
>        struct list_head unused;        /* List of unused blocks */
>        struct list_head used;          /* List of used blocks */
> +       struct mutex lock;
>
>        /* Map host addr (kmalloced/vmalloced) to device logical addr. */
>        struct ct_vm_block *(*map)(struct ct_vm *, void *host_addr, int
> size);
>

I tried this patch (actually I downloaded the snapshot yesterday and patch
said it was already applied there), but my system can not boot at all with
this driver. It prints 'starting udev' at that's it. I removed the driver
and my system is back to normal.

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: snd-ctxfi
  2009-06-04  3:48           ` snd-ctxfi The Source
@ 2009-06-04  5:29             ` Takashi Iwai
  0 siblings, 0 replies; 8+ messages in thread
From: Takashi Iwai @ 2009-06-04  5:29 UTC (permalink / raw)
  To: The Source; +Cc: alsa-devel

At Thu, 4 Jun 2009 07:48:47 +0400,
The Source wrote:
> 
> 2009/6/2 Takashi Iwai <tiwai@suse.de>
> 
>     At Mon, 1 Jun 2009 21:47:02 -0400,
>     Lee Revell wrote:
>     >
>     > On Mon, Jun 1, 2009 at 5:10 AM, Takashi Iwai <tiwai@suse.de> wrote:
>     > > Thanks.  So, something wrong in the mmap handling.
>     > > I'll take a look later in this week.
>     >
>     > Looks like ct_map_audio_buffer calls ct_vm_map under spinlock;
>     > ct_vm_map calls get_vm_block which calls kzalloc with GFP_KERNEL.
>    
>     Indeed.  Does the patch below fix the problem?
> 
>     thanks,
>    
>     Takashi
>    
>     ---
>     diff --git a/sound/pci/ctxfi/ctatc.c b/sound/pci/ctxfi/ctatc.c
>     index ead104e..1a4bb35 100644
>     --- a/sound/pci/ctxfi/ctatc.c
>     +++ b/sound/pci/ctxfi/ctatc.c
>     @@ -119,7 +119,6 @@ atc_pcm_release_resources(struct ct_atc *atc, struct
>     ct_atc_pcm *apcm);
>    
>      static int ct_map_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm
>     *apcm)
>      {
>     -       unsigned long flags;
>            struct snd_pcm_runtime *runtime;
>            struct ct_vm *vm;
>    
>     @@ -129,9 +128,7 @@ static int ct_map_audio_buffer(struct ct_atc *atc,
>     struct ct_atc_pcm *apcm)
>            runtime = apcm->substream->runtime;
>            vm = atc->vm;
>    
>     -       spin_lock_irqsave(&atc->vm_lock, flags);
>            apcm->vm_block = vm->map(vm, runtime->dma_area, runtime->
>     dma_bytes);
>     -       spin_unlock_irqrestore(&atc->vm_lock, flags);
>    
>            if (NULL == apcm->vm_block)
>                    return -ENOENT;
>     @@ -141,7 +138,6 @@ static int ct_map_audio_buffer(struct ct_atc *atc,
>     struct ct_atc_pcm *apcm)
>    
>      static void ct_unmap_audio_buffer(struct ct_atc *atc, struct ct_atc_pcm
>     *apcm)
>      {
>     -       unsigned long flags;
>            struct ct_vm *vm;
>    
>            if (NULL == apcm->vm_block)
>     @@ -149,9 +145,7 @@ static void ct_unmap_audio_buffer(struct ct_atc *atc,
>     struct ct_atc_pcm *apcm)
>    
>            vm = atc->vm;
>    
>     -       spin_lock_irqsave(&atc->vm_lock, flags);
>            vm->unmap(vm, apcm->vm_block);
>     -       spin_unlock_irqrestore(&atc->vm_lock, flags);
>    
>            apcm->vm_block = NULL;
>      }
>     @@ -161,9 +155,7 @@ static unsigned long atc_get_ptp_phys(struct ct_atc
>     *atc, int index)
>            struct ct_vm *vm;
>            void *kvirt_addr;
>            unsigned long phys_addr;
>     -       unsigned long flags;
>    
>     -       spin_lock_irqsave(&atc->vm_lock, flags);
>            vm = atc->vm;
>            kvirt_addr = vm->get_ptp_virt(vm, index);
>            if (kvirt_addr == NULL)
>     @@ -171,8 +163,6 @@ static unsigned long atc_get_ptp_phys(struct ct_atc
>     *atc, int index)
>            else
>                    phys_addr = virt_to_phys(kvirt_addr);
>    
>     -       spin_unlock_irqrestore(&atc->vm_lock, flags);
>     -
>            return phys_addr;
>      }
>    
>     @@ -1562,7 +1552,6 @@ int ct_atc_create(struct snd_card *card, struct
>     pci_dev *pci,
>            atc_set_ops(atc);
>    
>            spin_lock_init(&atc->atc_lock);
>     -       spin_lock_init(&atc->vm_lock);
>    
>            /* Find card model */
>            err = atc_identify_card(atc);
>     diff --git a/sound/pci/ctxfi/ctatc.h b/sound/pci/ctxfi/ctatc.h
>     index 286c993..a7b0ec2 100644
>     --- a/sound/pci/ctxfi/ctatc.h
>     +++ b/sound/pci/ctxfi/ctatc.h
>     @@ -101,7 +101,6 @@ struct ct_atc {
>            unsigned long (*get_ptp_phys)(struct ct_atc *atc, int index);
>    
>            spinlock_t atc_lock;
>     -       spinlock_t vm_lock;
>    
>            int (*pcm_playback_prepare)(struct ct_atc *atc,
>                                        struct ct_atc_pcm *apcm);
>     diff --git a/sound/pci/ctxfi/ctvmem.c b/sound/pci/ctxfi/ctvmem.c
>     index cecf77e..363b67e 100644
>     --- a/sound/pci/ctxfi/ctvmem.c
>     +++ b/sound/pci/ctxfi/ctvmem.c
>     @@ -35,25 +35,27 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
>            struct ct_vm_block *block = NULL, *entry = NULL;
>            struct list_head *pos = NULL;
>    
>     +       mutex_lock(&vm->lock);
>            list_for_each(pos, &vm->unused) {
>                    entry = list_entry(pos, struct ct_vm_block, list);
>                    if (entry->size >= size)
>                            break; /* found a block that is big enough */
>            }
>            if (pos == &vm->unused)
>     -               return NULL;
>     +               goto out;
>    
>            if (entry->size == size) {
>                    /* Move the vm node from unused list to used list directly
>     */
>                    list_del(&entry->list);
>                    list_add(&entry->list, &vm->used);
>                    vm->size -= size;
>     -               return entry;
>     +               block = entry;
>     +               goto out;
>            }
>    
>            block = kzalloc(sizeof(*block), GFP_KERNEL);
>            if (NULL == block)
>     -               return NULL;
>     +               goto out;
>    
>            block->addr = entry->addr;
>            block->size = size;
>     @@ -62,6 +64,8 @@ get_vm_block(struct ct_vm *vm, unsigned int size)
>            entry->size -= size;
>            vm->size -= size;
>    
>     + out:
>     +       mutex_unlock(&vm->lock);
>            return block;
>      }
>    
>     @@ -70,6 +74,7 @@ static void put_vm_block(struct ct_vm *vm, struct
>     ct_vm_block *block)
>            struct ct_vm_block *entry = NULL, *pre_ent = NULL;
>            struct list_head *pos = NULL, *pre = NULL;
>    
>     +       mutex_lock(&vm->lock);
>            list_del(&block->list);
>            vm->size += block->size;
>    
>     @@ -106,6 +111,7 @@ static void put_vm_block(struct ct_vm *vm, struct
>     ct_vm_block *block)
>                    pos = pre;
>                    pre = pos->prev;
>            }
>     +       mutex_unlock(&vm->lock);
>      }
>    
>      /* Map host addr (kmalloced/vmalloced) to device logical addr. */
>     @@ -191,6 +197,8 @@ int ct_vm_create(struct ct_vm **rvm)
>            if (NULL == vm)
>                    return -ENOMEM;
>    
>     +       mutex_init(&vm->lock);
>     +
>            /* Allocate page table pages */
>            for (i = 0; i < CT_PTP_NUM; i++) {
>                    vm->ptp[i] = kmalloc(PAGE_SIZE, GFP_KERNEL);
>     diff --git a/sound/pci/ctxfi/ctvmem.h b/sound/pci/ctxfi/ctvmem.h
>     index 4eb5bdd..618952e 100644
>     --- a/sound/pci/ctxfi/ctvmem.h
>     +++ b/sound/pci/ctxfi/ctvmem.h
>     @@ -20,7 +20,7 @@
>    
>      #define CT_PTP_NUM     1       /* num of device page table pages */
>    
>     -#include <linux/spinlock.h>
>     +#include <linux/mutex.h>
>      #include <linux/list.h>
>    
>      struct ct_vm_block {
>     @@ -35,6 +35,7 @@ struct ct_vm {
>            unsigned int size;              /* Available addr space in bytes */
>            struct list_head unused;        /* List of unused blocks */
>            struct list_head used;          /* List of used blocks */
>     +       struct mutex lock;
>    
>            /* Map host addr (kmalloced/vmalloced) to device logical addr. */
>            struct ct_vm_block *(*map)(struct ct_vm *, void *host_addr, int
>     size);
> 
> I tried this patch (actually I downloaded the snapshot yesterday and patch
> said it was already applied there), but my system can not boot at all with
> this driver. It prints 'starting udev' at that's it. I removed the driver and
> my system is back to normal.

Yes, yesterday's version had a severe bug in the core code.
It's already fixed, so please use the later one.

Note that alsa-driver-snapshot.tar.gz is *always* the latest version,
so pick this up instead of the daily snapshots for testing something.


thanks,

Takashi

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2009-06-04  5:29 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-30  3:22 snd-ctxfi The Source
2009-05-30  7:00 ` snd-ctxfi Takashi Iwai
2009-05-30 16:22   ` snd-ctxfi The Source
2009-06-01  9:10     ` snd-ctxfi Takashi Iwai
2009-06-02  1:47       ` snd-ctxfi Lee Revell
2009-06-02  6:09         ` snd-ctxfi Takashi Iwai
2009-06-04  3:48           ` snd-ctxfi The Source
2009-06-04  5:29             ` snd-ctxfi Takashi Iwai

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.