linux-fbdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Aurelien Jacobs <aurel@gnuage.org>
To: linux-fbdev-devel@lists.sourceforge.net, Andrew Morton <akpm@osdl.org>
Subject: Re: Re: [PATCH] vesafb memory size mismatch
Date: Sun, 3 Oct 2004 02:53:23 +0200	[thread overview]
Message-ID: <20041003025323.11efa1b5@homer.gnuage.org> (raw)
In-Reply-To: <87acv6l9ru.fsf@bytesex.org>

[-- Attachment #1: Type: text/plain, Size: 1013 bytes --]

On 01 Oct 2004 17:48:21 +0200
Gerd Knorr <kraxel@bytesex.org> wrote:

> Aurelien Jacobs <aurel@gnuage.org> writes:
> 
> > Hi,
> > 
> > I found a bug in the vesafb driver. The video memory size which is
> > reserved by vesafb do not correspond to the really needed memory.
> 
> Oh, while we are at it, I've hacked a patch as well some time ago.
> All the size calculation in vesafb is a bit tricky.  I've tried to
> cleanup that a bit and fix some bugs along the way, but never managed
> to submit it.  I think the boundary check cleanups should also fix the
> bug you've seen.

Unfortunately not :-(
Your patch still keep the same way to calculate minimum memory
required for the video mode (based on xres*bytes_per_pixel). It's
wrong ! With video cards such as 3dfx in 24 bits mode, line_length
is bigger than xres*bytes_per_pixel (probably for alignment pupose).
So not enough memory is allocated, leading to a crash.

Here is an updated version of your patch with the right way to
calculate size_vmode.

Aurel

[-- Attachment #2: vesafb-memory.diff --]
[-- Type: text/plain, Size: 3840 bytes --]

diff -aur linux-2.6.8.1.orig/drivers/video/vesafb.c linux-2.6.8.1/drivers/video/vesafb.c
--- linux-2.6.8.1.orig/drivers/video/vesafb.c	2004-10-03 02:34:53.652872688 +0200
+++ linux-2.6.8.1/drivers/video/vesafb.c	2004-10-03 02:39:23.401864608 +0200
@@ -218,6 +218,9 @@
 	struct platform_device *dev = to_platform_device(device);
 	struct fb_info *info;
 	int i, err;
+	unsigned int size_vmode;
+	unsigned int size_remap;
+	unsigned int size_total;
 
 	if (screen_info.orig_video_isVGA != VIDEO_TYPE_VLFB)
 		return -ENXIO;
@@ -229,32 +232,36 @@
 	vesafb_defined.xres = screen_info.lfb_width;
 	vesafb_defined.yres = screen_info.lfb_height;
 	vesafb_fix.line_length = screen_info.lfb_linelength;
-
-	/* Allocate enough memory for double buffering */
-	vesafb_fix.smem_len = screen_info.lfb_width * screen_info.lfb_height * vesafb_defined.bits_per_pixel >> 2;
-
-	/* check that we don't remap more memory than old cards have */
-	if (vesafb_fix.smem_len > (screen_info.lfb_size * 65536))
-		vesafb_fix.smem_len = screen_info.lfb_size * 65536;
-
-	/* Set video size according to vram boot option */
-	if (vram)
-		vesafb_fix.smem_len = vram * 1024 * 1024;
-
 	vesafb_fix.visual   = (vesafb_defined.bits_per_pixel == 8) ?
 		FB_VISUAL_PSEUDOCOLOR : FB_VISUAL_TRUECOLOR;
 
-	/* limit framebuffer size to 16 MB.  Otherwise we'll eat tons of
-	 * kernel address space for nothing if the gfx card has alot of
-	 * memory (>= 128 MB isn't uncommon these days ...) */
-	if (vesafb_fix.smem_len > 16 * 1024 * 1024)
-		vesafb_fix.smem_len = 16 * 1024 * 1024;
+	/*   size_vmode -- that is the amount of memory needed for the
+	 *                 used video mode, i.e. the minimum amount of
+	 *                 memory we need. */
+	size_vmode = vesafb_fix.line_length * vesafb_defined.yres;
+
+	/*   size_total -- all video memory we have. Used for mtrr
+	 *                 entries and bounds checking. */
+	size_total = screen_info.lfb_size * 65536;
+	if (size_total < size_vmode)
+		size_total = size_vmode;
+	if (vram)
+		size_total = vram * 1024 * 1024;
+
+	/*   size_remap -- the amount of video memory we are going to
+	 *                 use for vesafb.  With modern cards it is no
+	 *                 option to simply use size_total as that
+	 *                 wastes plenty of kernel address space. */
+	size_remap  = size_vmode * 2;
+	if (size_remap > size_total)
+		size_remap = size_total;
+	vesafb_fix.smem_len = size_remap;
 
 #ifndef __i386__
 	screen_info.vesapm_seg = 0;
 #endif
 
-	if (!request_mem_region(vesafb_fix.smem_start, vesafb_fix.smem_len, "vesafb")) {
+	if (!request_mem_region(vesafb_fix.smem_start, size_total, "vesafb")) {
 		printk(KERN_WARNING
 		       "vesafb: abort, cannot reserve video memory at 0x%lx\n",
 			vesafb_fix.smem_start);
@@ -279,8 +286,10 @@
 		goto err;
 	}
 
-	printk(KERN_INFO "vesafb: framebuffer at 0x%lx, mapped to 0x%p, size %dk\n",
-	       vesafb_fix.smem_start, info->screen_base, vesafb_fix.smem_len/1024);
+	printk(KERN_INFO "vesafb: framebuffer at 0x%lx, mapped to 0x%p, "
+	       "using %dk, total %dk\n",
+	       vesafb_fix.smem_start, info->screen_base,
+	       size_remap/1024, size_total/1024);
 	printk(KERN_INFO "vesafb: mode is %dx%dx%d, linelength=%d, pages=%d\n",
 	       vesafb_defined.xres, vesafb_defined.yres, vesafb_defined.bits_per_pixel, vesafb_fix.line_length, screen_info.pages);
 
@@ -364,7 +373,7 @@
 	request_region(0x3c0, 32, "vesafb");
 
 	if (mtrr) {
-		int temp_size = vesafb_fix.smem_len;
+		int temp_size = size_total;
 		/* Find the largest power-of-two */
 		while (temp_size & (temp_size - 1))
                 	temp_size &= (temp_size - 1);
@@ -395,7 +404,7 @@
 	return 0;
 err:
 	framebuffer_release(info);
-	release_mem_region(vesafb_fix.smem_start, vesafb_fix.smem_len);
+	release_mem_region(vesafb_fix.smem_start, size_total);
 	return err;
 }
 

  parent reply	other threads:[~2004-10-03  1:25 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-10-01 13:36 [PATCH] vesafb memory size mismatch Aurelien Jacobs
2004-10-01 15:48 ` Gerd Knorr
2004-10-01 20:22   ` [Linux-fbdev-devel] " Antonino A. Daplas
2004-10-04  9:27     ` Gerd Knorr
2004-10-03  0:53   ` Aurelien Jacobs [this message]
2004-10-03  1:59     ` Andrew Morton
2004-10-03  8:46       ` Antonino A. Daplas

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20041003025323.11efa1b5@homer.gnuage.org \
    --to=aurel@gnuage.org \
    --cc=akpm@osdl.org \
    --cc=linux-fbdev-devel@lists.sourceforge.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).