Averatec 6240 pcmcia_socket0: unable to apply power

[Daniel Ritz]

On Thursday 08 September 2005 19.33, David Hinds wrote:
> On Thu, Sep 08, 2005 at 06:30:23PM +0200, Daniel Ritz wrote:
> > hi james
> > 
> > [ please always use reply-to-all. i'm not subscribed to the list ]
> > 
> > you have a notebook with graphic controller working in shared memory mode
> > which means it uses the main memory. unfortunatley the is BIOS crap and
> > forgets to report that. add the following to the kernel command line to
> > fix the problem:
> > 	reserve=0x1e000000,0x2000000
> > it tells the kernel not to use the region from 480-512 MB.
> 
> I thought that this had been fixed, so the kernel would round up
> apparent memory sizes that were a bit smaller than a power of two.

it currently only rounds up to the next MB. but that doesn't catch
the shared VGAs...

> I guess not.  It really should get fixed properly (even though it
> technically is not a kernel problem), because it generates a huge
> amount of problem reports.
> 

something like the attached patch...

cc:ing linus as he is the one that added the "find the largest gap" code
that comes handy here...
and cc:ing andrew for the obvious reason...

rgds
-daniel

----

[PATCH] i386/x86_64: align pci iomem start address to 64MB

align the PCI iomemory starting address to a muliple of 64MB to avoid problems
with some BIOS that forget to report the memory used by a shared memory graphic
controller in the e820 memory map.

most people don't give the shared VGA more than 64MB, right?

Signed-off-by: Daniel Ritz <daniel.ritz at gmx.ch>

diff --git a/arch/i386/kernel/setup.c b/arch/i386/kernel/setup.c
--- a/arch/i386/kernel/setup.c
+++ b/arch/i386/kernel/setup.c
@@ -1325,7 +1325,16 @@ static void __init register_memory(void)
 	i = e820.nr_map;
 	while (--i >= 0) {
 		unsigned long long start = e820.map[i].addr;
-		unsigned long long end = start + e820.map[i].size;
+		unsigned long long end;
+
+		/*
+		 * try to find the largest gap starting at a muliple of 64MB.
+		 * this is to avoid problems with some BIOSen that forget to report
+		 * memory used by the shared memory graphic controller.
+		 * most people are not going to use more than 64MB for a shared
+		 * controller, right? round up to 128MB?
+		 */
+		end = (start + e820.map[i].size + 0x3ffffff) & ~0x3ffffff;
 
 		/*
 		 * Since "last" is at most 4GB, we know we'll
@@ -1344,14 +1353,10 @@ static void __init register_memory(void)
 	}
 
 	/*
-	 * Start allocating dynamic PCI memory a bit into the gap,
-	 * aligned up to the nearest megabyte.
-	 *
-	 * Question: should we try to pad it up a bit (do something
-	 * like " + (gapsize >> 3)" in there too?). We now have the
-	 * technology.
+	 * Start allocating dynamic PCI memory at gapstart which is already
+	 * rounded up to a multiple of 6 MB.
 	 */
-	pci_mem_start = (gapstart + 0xfffff) & ~0xfffff;
+	pci_mem_start = gapstart;
 
 	printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
 		pci_mem_start, gapstart, gapsize);
diff --git a/arch/x86_64/kernel/e820.c b/arch/x86_64/kernel/e820.c
--- a/arch/x86_64/kernel/e820.c
+++ b/arch/x86_64/kernel/e820.c
@@ -578,7 +578,16 @@ __init void e820_setup_gap(void)
 	i = e820.nr_map;
 	while (--i >= 0) {
 		unsigned long long start = e820.map[i].addr;
-		unsigned long long end = start + e820.map[i].size;
+		unsigned long long end;
+
+		/*
+		 * try to find the largest gap starting at a muliple of 64MB.
+		 * this is to avoid problems with some BIOSen that forget to report
+		 * memory used by the shared memory graphic controller.
+		 * most people are not going to use more than 64MB for a shared
+		 * controller, right? round up to 128MB?
+		 */
+		end = (start + e820.map[i].size + 0x3ffffff) & ~0x3ffffff;
 
 		/*
 		 * Since "last" is at most 4GB, we know we'll
@@ -604,14 +613,10 @@ __init void e820_setup_gap(void)
 	}
 
 	/*
-	 * Start allocating dynamic PCI memory a bit into the gap,
-	 * aligned up to the nearest megabyte.
-	 *
-	 * Question: should we try to pad it up a bit (do something
-	 * like " + (gapsize >> 3)" in there too?). We now have the
-	 * technology.
+	 * Start allocating dynamic PCI memory at gapstart which is already
+	 * rounded up to a multiple of 6 MB.
 	 */
-	pci_mem_start = (gapstart + 0xfffff) & ~0xfffff;
+	pci_mem_start = gapstart;
 
 	printk(KERN_INFO "Allocating PCI resources starting at %lx (gap: %lx:%lx)\n",
 		pci_mem_start, gapstart, gapsize);