Re: "Cannot allocate memory" problem with latest CVS

[Takashi Iwai <tiwai@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 515 bytes --]

At Mon, 28 Oct 2002 13:21:48 +0000 (GMT),
Chris Rankin wrote:
> 
> Ouch. This means that this problem will always appear
> eventually, and the only way to clear it is waiting
> for Linux to free a large amount of cache memory.
> <flamebait>Or I could reboot my machine.</flamebait>
 
could you try the attached patch?
as mentioned in my previous mail, this is really a hack, but if it
works...


> In the meantime, I have no PCM playback via my
> SB-Live!. Is this a 2.4 or a SB-Live limitation?

both :)


Takashi

[-- Attachment #2: sgbuf-getpage.dif --]
[-- Type: application/octet-stream, Size: 1211 bytes --]

Index: alsa-kernel/core/pcm_sgbuf.c
===================================================================
RCS file: /suse/tiwai/cvs/alsa/alsa-kernel/core/pcm_sgbuf.c,v
retrieving revision 1.7
diff -u -r1.7 pcm_sgbuf.c
--- alsa-kernel/core/pcm_sgbuf.c	10 Sep 2002 10:20:00 -0000	1.7
+++ alsa-kernel/core/pcm_sgbuf.c	28 Oct 2002 14:35:43 -0000
@@ -128,7 +128,32 @@
 	while (sgbuf->pages < pages) {
 		void *ptr;
 		dma_addr_t addr;
+#ifdef __i386__
+		/* HACK! we try to allocate with GFP_KERNEL */
+		unsigned long rmask;
+		if (sgbuf->pci)
+			rmask = ~(unsigned long)sgbuf->pci->dma_mask;
+		else
+			rmask = 0;
+		ptr = (void *)__get_free_pages(GFP_KERNEL, 1);
+		if (ptr) {
+			addr = virt_to_phys(ptr);
+			if (((unsigned long)addr + PAGE_SIZE - 1) & rmask) {
+				/* reallocate with the proper mask */
+				free_pages((unsigned long)ptr, PAGE_SIZE);
+				ptr = (void *)__get_free_pages(GFP_KERNEL | GFP_DMA, 1);
+				if (ptr)
+					addr = virt_to_phys(ptr);
+				else
+					addr = 0;
+			}
+		} else
+			addr = 0;
+		if (ptr)
+			memset(ptr, 0, PAGE_SIZE);
+#else
 		ptr = snd_malloc_pci_pages(sgbuf->pci, PAGE_SIZE, &addr);
+#endif
 		if (! ptr)
 			return -ENOMEM;
 		sgbuf->table[sgbuf->pages].buf = ptr;