aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorHans Verkuil <hverkuil@xs4all.nl>2009-02-18 13:20:05 -0300
committerMauro Carvalho Chehab <mchehab@redhat.com>2009-03-30 12:42:55 -0300
commitb80696b700b25b7aa054610af84a1f05e4f4ea58 (patch)
tree375e1654ca11cabf4a263d8486e8fc3450b9d162
parent96b8e145f0a6f3c0470b5cdbe7ba9be314bb556f (diff)
V4L/DVB (10704): zoran: remove broken BIGPHYS_AREA and BUZ_HIMEM code, and allow for kmallocs > 128 kB
Remove memory allocation madness. Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
-rw-r--r--drivers/media/video/zoran/zoran.h2
-rw-r--r--drivers/media/video/zoran/zoran_driver.c189
2 files changed, 31 insertions, 160 deletions
diff --git a/drivers/media/video/zoran/zoran.h b/drivers/media/video/zoran/zoran.h
index 1bf540cb546..76668bda049 100644
--- a/drivers/media/video/zoran/zoran.h
+++ b/drivers/media/video/zoran/zoran.h
@@ -170,8 +170,6 @@ Private IOCTL to set up for displaying MJPEG
#endif
#define V4L_MASK_FRAME (V4L_MAX_FRAME - 1)
-#define MAX_KMALLOC_MEM (128*1024)
-
#include "zr36057.h"
enum card_type {
diff --git a/drivers/media/video/zoran/zoran_driver.c b/drivers/media/video/zoran/zoran_driver.c
index dc6ba554597..1f8d3a8b0cb 100644
--- a/drivers/media/video/zoran/zoran_driver.c
+++ b/drivers/media/video/zoran/zoran_driver.c
@@ -191,9 +191,6 @@ const struct zoran_format zoran_formats[] = {
};
#define NUM_FORMATS ARRAY_SIZE(zoran_formats)
-// RJ: Test only - want to test BUZ_USE_HIMEM even when CONFIG_BIGPHYS_AREA is defined
-
-
static int lock_norm; /* 0 = default 1 = Don't change TV standard (norm) */
module_param(lock_norm, int, 0644);
MODULE_PARM_DESC(lock_norm, "Prevent norm changes (1 = ignore, >1 = fail)");
@@ -229,77 +226,8 @@ static void jpg_fbuffer_free(struct file *file);
* Allocate the V4L grab buffers
*
* These have to be pysically contiguous.
- * If v4l_bufsize <= MAX_KMALLOC_MEM we use kmalloc
- * else we try to allocate them with bigphysarea_alloc_pages
- * if the bigphysarea patch is present in the kernel,
- * else we try to use high memory (if the user has bootet
- * Linux with the necessary memory left over).
- */
-
-static unsigned long
-get_high_mem (unsigned long size)
-{
-/*
- * Check if there is usable memory at the end of Linux memory
- * of at least size. Return the physical address of this memory,
- * return 0 on failure.
- *
- * The idea is from Alexandro Rubini's book "Linux device drivers".
- * The driver from him which is downloadable from O'Reilly's
- * web site misses the "virt_to_phys(high_memory)" part
- * (and therefore doesn't work at all - at least with 2.2.x kernels).
- *
- * It should be unnecessary to mention that THIS IS DANGEROUS,
- * if more than one driver at a time has the idea to use this memory!!!!
*/
- volatile unsigned char __iomem *mem;
- unsigned char c;
- unsigned long hi_mem_ph;
- unsigned long i;
-
- /* Map the high memory to user space */
-
- hi_mem_ph = virt_to_phys(high_memory);
-
- mem = ioremap(hi_mem_ph, size);
- if (!mem) {
- dprintk(1,
- KERN_ERR "%s: get_high_mem() - ioremap failed\n",
- ZORAN_NAME);
- return 0;
- }
-
- for (i = 0; i < size; i++) {
- /* Check if it is memory */
- c = i & 0xff;
- writeb(c, mem + i);
- if (readb(mem + i) != c)
- break;
- c = 255 - c;
- writeb(c, mem + i);
- if (readb(mem + i) != c)
- break;
- writeb(0, mem + i); /* zero out memory */
-
- /* give the kernel air to breath */
- if ((i & 0x3ffff) == 0x3ffff)
- schedule();
- }
-
- iounmap(mem);
-
- if (i != size) {
- dprintk(1,
- KERN_ERR
- "%s: get_high_mem() - requested %lu, avail %lu\n",
- ZORAN_NAME, size, i);
- return 0;
- }
-
- return hi_mem_ph;
-}
-
static int
v4l_fbuffer_alloc (struct file *file)
{
@@ -307,82 +235,37 @@ v4l_fbuffer_alloc (struct file *file)
struct zoran *zr = fh->zr;
int i, off;
unsigned char *mem;
- unsigned long pmem = 0;
for (i = 0; i < fh->v4l_buffers.num_buffers; i++) {
if (fh->v4l_buffers.buffer[i].fbuffer)
dprintk(2,
KERN_WARNING
- "%s: v4l_fbuffer_alloc() - buffer %d allready allocated!?\n",
+ "%s: v4l_fbuffer_alloc() - buffer %d already allocated!?\n",
ZR_DEVNAME(zr), i);
//udelay(20);
- if (fh->v4l_buffers.buffer_size <= MAX_KMALLOC_MEM) {
- /* Use kmalloc */
-
- mem = kmalloc(fh->v4l_buffers.buffer_size, GFP_KERNEL);
- if (!mem) {
- dprintk(1,
- KERN_ERR
- "%s: v4l_fbuffer_alloc() - kmalloc for V4L buf %d failed\n",
- ZR_DEVNAME(zr), i);
- v4l_fbuffer_free(file);
- return -ENOBUFS;
- }
- fh->v4l_buffers.buffer[i].fbuffer = mem;
- fh->v4l_buffers.buffer[i].fbuffer_phys =
- virt_to_phys(mem);
- fh->v4l_buffers.buffer[i].fbuffer_bus =
- virt_to_bus(mem);
- for (off = 0; off < fh->v4l_buffers.buffer_size;
- off += PAGE_SIZE)
- SetPageReserved(MAP_NR(mem + off));
- dprintk(4,
- KERN_INFO
- "%s: v4l_fbuffer_alloc() - V4L frame %d mem 0x%lx (bus: 0x%llx)\n",
- ZR_DEVNAME(zr), i, (unsigned long) mem,
- (unsigned long long)virt_to_bus(mem));
- } else {
-
- /* Use high memory which has been left at boot time */
-
- /* Ok., Ok. this is an evil hack - we make
- * the assumption that physical addresses are
- * the same as bus addresses (true at least
- * for Intel processors). The whole method of
- * obtaining and using this memory is not very
- * nice - but I hope it saves some poor users
- * from kernel hacking, which might have even
- * more evil results */
-
- if (i == 0) {
- int size =
- fh->v4l_buffers.num_buffers *
- fh->v4l_buffers.buffer_size;
-
- pmem = get_high_mem(size);
- if (pmem == 0) {
- dprintk(1,
- KERN_ERR
- "%s: v4l_fbuffer_alloc() - get_high_mem (size = %d KB) for V4L bufs failed\n",
- ZR_DEVNAME(zr), size >> 10);
- return -ENOBUFS;
- }
- fh->v4l_buffers.buffer[0].fbuffer = NULL;
- fh->v4l_buffers.buffer[0].fbuffer_phys = pmem;
- fh->v4l_buffers.buffer[0].fbuffer_bus = pmem;
- dprintk(4,
- KERN_INFO
- "%s: v4l_fbuffer_alloc() - using %d KB high memory\n",
- ZR_DEVNAME(zr), size >> 10);
- } else {
- fh->v4l_buffers.buffer[i].fbuffer = NULL;
- fh->v4l_buffers.buffer[i].fbuffer_phys =
- pmem + i * fh->v4l_buffers.buffer_size;
- fh->v4l_buffers.buffer[i].fbuffer_bus =
- pmem + i * fh->v4l_buffers.buffer_size;
- }
+ mem = kmalloc(fh->v4l_buffers.buffer_size, GFP_KERNEL);
+ if (!mem) {
+ dprintk(1,
+ KERN_ERR
+ "%s: v4l_fbuffer_alloc() - kmalloc for V4L buf %d failed\n",
+ ZR_DEVNAME(zr), i);
+ v4l_fbuffer_free(file);
+ return -ENOBUFS;
}
+ fh->v4l_buffers.buffer[i].fbuffer = mem;
+ fh->v4l_buffers.buffer[i].fbuffer_phys =
+ virt_to_phys(mem);
+ fh->v4l_buffers.buffer[i].fbuffer_bus =
+ virt_to_bus(mem);
+ for (off = 0; off < fh->v4l_buffers.buffer_size;
+ off += PAGE_SIZE)
+ SetPageReserved(MAP_NR(mem + off));
+ dprintk(4,
+ KERN_INFO
+ "%s: v4l_fbuffer_alloc() - V4L frame %d mem 0x%lx (bus: 0x%lx)\n",
+ ZR_DEVNAME(zr), i, (unsigned long) mem,
+ virt_to_bus(mem));
}
fh->v4l_buffers.allocated = 1;
@@ -405,13 +288,11 @@ v4l_fbuffer_free (struct file *file)
if (!fh->v4l_buffers.buffer[i].fbuffer)
continue;
- if (fh->v4l_buffers.buffer_size <= MAX_KMALLOC_MEM) {
- mem = fh->v4l_buffers.buffer[i].fbuffer;
- for (off = 0; off < fh->v4l_buffers.buffer_size;
- off += PAGE_SIZE)
- ClearPageReserved(MAP_NR(mem + off));
- kfree((void *) fh->v4l_buffers.buffer[i].fbuffer);
- }
+ mem = fh->v4l_buffers.buffer[i].fbuffer;
+ for (off = 0; off < fh->v4l_buffers.buffer_size;
+ off += PAGE_SIZE)
+ ClearPageReserved(MAP_NR(mem + off));
+ kfree((void *) fh->v4l_buffers.buffer[i].fbuffer);
fh->v4l_buffers.buffer[i].fbuffer = NULL;
}
@@ -421,16 +302,10 @@ v4l_fbuffer_free (struct file *file)
/*
* Allocate the MJPEG grab buffers.
*
- * If the requested buffer size is smaller than MAX_KMALLOC_MEM,
- * kmalloc is used to request a physically contiguous area,
- * else we allocate the memory in framgents with get_zeroed_page.
- *
* If a Natoma chipset is present and this is a revision 1 zr36057,
* each MJPEG buffer needs to be physically contiguous.
* (RJ: This statement is from Dave Perks' original driver,
* I could never check it because I have a zr36067)
- * The driver cares about this because it reduces the buffer
- * size to MAX_KMALLOC_MEM in that case (which forces contiguous allocation).
*
* RJ: The contents grab buffers needs never be accessed in the driver.
* Therefore there is no need to allocate them with vmalloc in order
@@ -464,7 +339,7 @@ jpg_fbuffer_alloc (struct file *file)
if (fh->jpg_buffers.buffer[i].frag_tab)
dprintk(2,
KERN_WARNING
- "%s: jpg_fbuffer_alloc() - buffer %d allready allocated!?\n",
+ "%s: jpg_fbuffer_alloc() - buffer %d already allocated!?\n",
ZR_DEVNAME(zr), i);
/* Allocate fragment table for this buffer */
@@ -504,7 +379,7 @@ jpg_fbuffer_alloc (struct file *file)
off += PAGE_SIZE)
SetPageReserved(MAP_NR(mem + off));
} else {
- /* jpg_bufsize is allreay page aligned */
+ /* jpg_bufsize is already page aligned */
for (j = 0;
j < fh->jpg_buffers.buffer_size / PAGE_SIZE;
j++) {
@@ -2173,6 +2048,7 @@ schan_unlock_and_return:
return res;
}
+
case VIDIOCGMBUF:
{
struct video_mbuf *vmbuf = arg;
@@ -2369,16 +2245,13 @@ sparams_unlock_and_return:
* tables to a Maximum of 2 MB */
if (breq->size > jpg_bufsize)
breq->size = jpg_bufsize;
- if (fh->jpg_buffers.need_contiguous &&
- breq->size > MAX_KMALLOC_MEM)
- breq->size = MAX_KMALLOC_MEM;
mutex_lock(&zr->resource_lock);
if (fh->jpg_buffers.allocated || fh->v4l_buffers.allocated) {
dprintk(1,
KERN_ERR
- "%s: BUZIOC_REQBUFS - buffers allready allocated\n",
+ "%s: BUZIOC_REQBUFS - buffers already allocated\n",
ZR_DEVNAME(zr));
res = -EBUSY;
goto jpgreqbuf_unlock_and_return;