Windows can't habdle large images

Main Page | Report this Page

Linux Forum Index » Linux Advocacy » Windows can't habdle large images

Page 3 of 10 Goto page Previous 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 Next

Windows can't habdle large images

Author

Message

Jim Richardson

Posted: Thu Jun 09, 2005 12:46 pm

Guest

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thu, 09 Jun 2005 09:17:32 -0600,
ray <ray@zianet.com> wrote:

Quote:

On Thu, 09 Jun 2005 09:35:21 +0200, Peter Köhlmann wrote:

Try this with XP (all patched)

HTML
BODY
IMG SRC="./sweetydead.jpg" width="9999999" height="9999999"
/BODY
/HTML

The image does not need to be that large, a very small one is sufficient
Just the /declaration/ of the size is enough to trigger the bug

Save your data before. The machine will BSOD

That seems to be a rather pathological example - I see little real
practical implication.

if the bug is limited to that one expression, you might be correct. But
consider the underlying cause.

What seems to be happening (to me, and I am no programmer, so take this
with a grain of salt) is that MS-Windows is allocating memory, based on
how big the image is claimed to be, without actually checking it, or
checking how much memory it *should* allocate, given any other
constraints (low memory system, whatever)

What this says to me, is that the VM subsystem dropped the ball on
sanity checks, and either goes nuts trying to swap out enough stuff to
fill the malloc (or equiv) request, or wraps around and writes crap out
of bounds.

How likely is it that this bug can be triggered in some other way?

dunno, but it's something to consider.

Meanwhile, the idea of the OS, being brought down by some poor html is
amusing.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)

iD8DBQFCqI6Dd90bcYOAWPYRArV8AJ4ihepC+lxAlWcXlEU3fm38QZ2xgQCgrZqL
iib6avs9VS5L4dAqm9o6kts=
=52RW
-----END PGP SIGNATURE-----

--
Jim Richardson http://www.eskimo.com/~warlock
You have the body of a 19 year old. Please return it before it gets wrinkled.

Peter Jensen

Posted: Thu Jun 09, 2005 12:57 pm

Guest

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Grug wrote:

Quote:

Wow... my machine came to a crawl...

Then I hit alt-f4 a few times and it was gone. whew.

Didn't BSOD, and it recovered quickly after I killed IE.

It might have done a BSOD if you had let it sit a while longer. I guess
you have to be on your toes when surfing from now on :-)

Quote:

And to say Windows can't handle large images, well...
9999999x9999999x32 (argb) pixels is about 3199999360 *gigabytes* of
memory.

Can't quite see how you got that figure. Every pixel is at most 4
bytes, which gives 9999999*9999999*4/1024^3 = 372529 GB. Still quite a
bit (no pun intended), though.

Quote:

I doubt many machines handle images that large Smile

My Mozilla here had no problem. The trick is of course to not do the
stupid thing and allocate memory for a 9999999x9999999 image, when you
just need a smaller image and then calculate the rest by stretching as
required.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCqJEsd1ZThqotgfgRAokMAKCroHOx95XRz5CI2XE1odO4HpWOmQCfVjcb
z65iaJ5NsS3qVUINqt0IROw=
=XXmR
-----END PGP SIGNATURE-----
--
PeKaJe

Pascal: A programming language named after a man who would turn over in
his grave if he knew about it. -- Datamation, January 15, 1984

Guest

Posted: Thu Jun 09, 2005 1:01 pm

Quote:

Oh, please.

Oh please what? It is a FACT that a driver bug was causing a kernel
panic in Linux. This bug would manifest itself when 'tar' was run and
bring down the system. FACT - not commentary or opinion.

So how exactly is a bad Windows driver any better or worse than a bad
Linux driver?

Tim Smith

Posted: Thu Jun 09, 2005 1:24 pm

Guest

In article <7-ednYAA2oJ98zXfRVn-sg@ctc.net>, robert wrote:

Quote:

The odds that everyone who has had this problem is using the same video driver
are probably pretty long. Did you test with multiple src's as suggested

How so? There are only a handful of video drivers nowadays.

--
--Tim Smith

DFS

Posted: Thu Jun 09, 2005 1:28 pm

Guest

jabailo@texeme.com wrote:

Quote:

Tom Shelton wrote:

It's a video driver error - it has nothing to do with memory. It
works on my little 64MB nVidia card that's in this box. Sure, it
slows things down, but it doesn't crash. The problem is crappy
video drivers - which, as we all know and have discussed, run in
kernel space. So, if the driver craps out, so does the system. It
appears that nVidia drivers (at least the latest) don't have this
problem.

Tom,

C'mon.

You windows guys have been rolling out this argument forever.

First you claim Linux doesn't have hardware support as an argument
against Linux.

Then everytime an exploit is pointed out you claim that it's "3rd
party software" that "isn't written right".

Bottom line: if an OS can't deal with incorrectly written application
software or drivers -- then it's not very good.

Then Linux stinks to high heaven. I recently installed an ATI 9600 card,
and when I next booted Linux it installed the driver and booted up fine and
I thought everything was working OK. It even seemed a bit snappier, and it
ran some of the [suck-ass] 3D games it previously wouldn't run with the
onboard video (Intel 865G Extreme Graphics).

But when I went into harddrake to change the screen resolution, it filled
the screen with mumbo-jumbo. The mouse was working, but there was nothing
but a mess on the screen. So I had to hardboot.

I'm sure some cola nut reader will blame me, or the ATI driver, or anything
but the OS.

This kind of thing is exactly why Linux is not ready for the desktop - and
probably never will be.

Linønut

Posted: Thu Jun 09, 2005 1:28 pm

Guest

lqualig@uku.co.uk poked his little head through the XP firewall and said:

Quote:

The point here is that a friggin WEB BROWSER has
the capability of potentially crashing a Windows system
simply because...

It's a driver issue.

Prove it.

Quote:

No different than the recent bug where sometimes
running 'tar' to archive a file causes a kernel panic in linux. Last I
checked 'tar' was even more rudimentary than a web browser so why is
tar bringing down a Linux system?

Never had that happen.

Ever.

--
When all you have is a hammer, everything looks like a nail.

Linønut

Posted: Thu Jun 09, 2005 1:29 pm

Guest

lqualig@uku.co.uk poked his little head through the XP firewall and said:

Quote:

Oh please what? It is a FACT that a driver bug was causing a kernel
panic in Linux. This bug would manifest itself when 'tar' was run and
bring down the system. FACT - not commentary or opinion.

URL, please. With a date.

Quote:

So how exactly is a bad Windows driver any better or worse than a bad
Linux driver?

Because we pay for Windows.

--
When all you have is a hammer, everything looks like a nail.

Tom Shelton

Posted: Thu Jun 09, 2005 1:34 pm

Guest

In article <42A88637.4050000@texeme.com>, jabailo@texeme.com wrote:

Quote:

Tom Shelton wrote:

It's a video driver error - it has nothing to do with memory. It works on
my little 64MB nVidia card that's in this box. Sure, it slows things
down, but it doesn't crash. The problem is crappy video drivers - which,
as we all know and have discussed, run in kernel space. So, if the driver
craps out, so does the system. It appears that nVidia drivers (at least
the latest) don't have this problem.

Tom,

C'mon.

You windows guys have been rolling out this argument forever.

First you claim Linux doesn't have hardware support as an argument
against Linux.

Linux has hardware support. And sometimes, it even has ok support. But,
other then my video card and printer, it's support hasn't been above ok.

1) My Kodak digital camera - works ok. gphoto2 is able to recognize it
and download pictures etc. Though, sometimes I have to plug it in a
couple of times. It also has a bad habbit of reseting the date and time
to it's default. On windows, the driver syncs the time to the system
time, and always recognizes the camera as soon as I push the button on
it's docking station.

2) My older hp digital. It no worky at all. To be fair, it barely
works under XP either - but it does work. That's why I when my friend
offered me an old card reader, I took it. See below :)

3) My card reader. This works ok on both - but it was slightly more
difficult to setup on linux. It's an older microtech CameraMate. On
Linux, I had to recompile my kernel (there is a kernel module to handle
this puppy). And, set it up in fstab. But, this leads to issues when I
want to use my muvo. I need to figure out how to write udev rules so it
can be more dynamic. On xp, I did have to find a driver - since it uses
a proprietary protocol. But, once installed I simply stick a card in
and go.

4) Muvo. Works well, but I haven't figured out udev well enough yet (my
fualt, I haven't spent the time on it yet) to be able to write rules so
that it can be hotswapped. Same issue with the card reader. Except
this one I didn't have to compile my kernel for. On xp - I just plug it
in and use it.

5) Scanner. HP5370C. Doesn't work on Linux at all. I had it close to
working before I went to a 2.6 kernel - at least apps would see it and
it would attempt to scan. But since I upgraded to 2.6 - nothing sees it
at all. On xp - no driver needed. It just works. Though, I do install
it's scanning software because it has more scanning options then the
default xp scan wizrd.

6) I have no complaints on my printer. It works well on both, though
the printer software on windows is a little nicer since it will show my
ink levels and such - but nothing to get worked up about, IMHO.

So, most of the hardware works - but I still have to figure out how to
make it hotplug. I don't have to figure that out on Windows. It just
happens. To be fair, the figure out part on Linux may have something to
do with the fact that I am using Gentoo. Gentoo, doesn't do a whole lot
for you like some of the more "user friendly" distributions. But, I
like it that way, it gives me a feel for how things "really" work.

Quote:

Then everytime an exploit is pointed out you claim that it's "3rd party
software" that "isn't written right".

Well... Most of the drivers are 3rd party. MS didn't write it - so if
it does bad things how is that MS's fault?

Quote:

Bottom line: if an OS can't deal with incorrectly written application
software or drivers -- then it's not very good.

I can't think of a single time I've had a user app crash 2K or XP... In
fact, like I said before - video drivers are the only driver I've ever
seen do this, and we all know why that is...

--
Tom Shelton

Linønut

Posted: Thu Jun 09, 2005 1:35 pm

Guest

Linønut poked his little head through the XP firewall and said:

Quote:

http://telon.dynalias.net/crashxp/crashxp.html

Mine is IE/SP2 on XP Pro. I'm viewing this page now. No crash, yet.
However, the whole computer is extremely unresponsive. Essentially
unusable. Ah, Task Manager finally came up. IE still won't respond to
user events. CPU usage is at 96%, then 2%, swinging wildly. Now the
mouse is frozen again. Ah, Ctrl-Alt-Del finally responded, now the
restart sequence is engaged. It's been going for a couple minutes now,
slowly ending apps one at a time. Finally, the boot screen.

On Linux (Firefox), no affect whatsoever.

--
When all you have is a hammer, everything looks like a nail.

Tom Shelton

Posted: Thu Jun 09, 2005 1:45 pm

Guest

In article <3dfmn2-8lb.ln1@grendel.myth>, Jim Richardson wrote:

Quote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thu, 09 Jun 2005 09:17:32 -0600,
ray <ray@zianet.com> wrote:
On Thu, 09 Jun 2005 09:35:21 +0200, Peter Köhlmann wrote:

Try this with XP (all patched)

HTML
BODY
IMG SRC="./sweetydead.jpg" width="9999999" height="9999999"
/BODY
/HTML

The image does not need to be that large, a very small one is sufficient
Just the /declaration/ of the size is enough to trigger the bug

Save your data before. The machine will BSOD

That seems to be a rather pathological example - I see little real
practical implication.

if the bug is limited to that one expression, you might be correct. But
consider the underlying cause.

What seems to be happening (to me, and I am no programmer, so take this
with a grain of salt) is that MS-Windows is allocating memory, based on
how big the image is claimed to be, without actually checking it, or
checking how much memory it *should* allocate, given any other
constraints (low memory system, whatever)

What this says to me, is that the VM subsystem dropped the ball on
sanity checks, and either goes nuts trying to swap out enough stuff to
fill the malloc (or equiv) request, or wraps around and writes crap out
of bounds.

How likely is it that this bug can be triggered in some other way?

dunno, but it's something to consider.

Meanwhile, the idea of the OS, being brought down by some poor html is
amusing.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)

iD8DBQFCqI6Dd90bcYOAWPYRArV8AJ4ihepC+lxAlWcXlEU3fm38QZ2xgQCgrZqL
iib6avs9VS5L4dAqm9o6kts=
=52RW
-----END PGP SIGNATURE-----

I think, Jim your ignoring the fact that some of us don't have the
system crash. I can load this just fine. Sure, it thrashes disk - but
once loaded, everything is fine. It scrolls slow - but it works. I've
never seen XP crash if it runs out of swap. I have seen it try to
allocate more and fail - but it has always just thrown an out of memory
error and killed the app.

This, if anything is most likely a display driver issue. And that is what
is causing the BSOD. Look at the crash screen shot. It was a fault in the
video driver.

You can of course, be amused that the video driver brought down the
system - but this is a well known design "descission" on windows. One,
I partly agree with. I personally think the driver should optionally
run in user space.

--
Tom Shelton

Guest

Posted: Thu Jun 09, 2005 1:45 pm

Quote:

What seems to be happening (to me, and
I am no programmer, so take this with
a grain of salt) is that MS-Windows
is allocating memory, based on how
big the image is claimed to be,
without actually checking it, or
checking how much memory it *should*
allocate, given any other constraints
(low memory system, whatever)

Quote:

What this says to me, is that the
VM subsystem dropped the ball on
sanity checks, and either goes nuts
trying to swap out enough stuff to
fill the malloc (or equiv) request,
or wraps around and writes crap out
of bounds.

Not bad for a non-programmer so let's look at this through a
programmers perspective.

"MS-Windows is allocating memory, based on how big the image is claimed
to be, without actually checking it" -

Keep in mind that Windows is the operating system. It will allocate as
much memory as the application (which is a browser in this case) asks
it to allocate. "Windows" itself doesn't know about HTML tags and how
large an image is claimed to be. "Windows" itself does not know where
this image is or how large it's supposed to be.

"What this says to me, is that the VM subsystem dropped the ball on
sanity checks..." -

By default the memory subsystem will attempt to satisfy all legitimate
memory requests. If the memory cannot be allocated then it will fail
and return a NULL pointer. You can write a simple app that loops and
keeps allocating memory. Eventually the free memory pool will be
exhausted and NULL will get returned. This isn't enough to crash the
OS.

The likely cause of the problem is something closer to this...

Internet Explorer retrieves the image it needs to display. The image
may be 320x200, 2000,2000 or whatever size you used. Once the image is
retrieved IE is able to determine its size. Lets assume it is 320x200
pixels.

IE then looks at the HTML tags and realizes that it needs to resize
this image to 9999999 x 9999999.

IE doesn't bother verifying if the size is reasonable or not or perhaps
just assumes that this will be handled at the next layer. In order to
scale an image on Windows a call to the StrechBlt() function is called.
This takes a source image, source size, and how large to draw the
destination image. So IE calls StretchBlt() and in plain-speak says
"Here's a 320x200 image. I need you to draw this image in my window but
scale it to 999999 x 999999."

The StretchBlt() function is usually handled by the video driver. The
video driver (which runs in kernel mode) does whatever it has to in
order to satisfy the request. For some video cards the onboard GPU may
be able to handle the resizing. Other video cards/drivers will handle
resizing in their own way.

Some drivers may need to allocate memory to do the resizing. If they
allocate a unreasonable amount of memory then the allocation request
will fail. A good video driver will realize that it didn't get the
memory it requested and return a failure code. A lesser quality video
driver will always "assume" the memory allocation succeeded and will go
ahead and use the memory anyway. In this case a kernel-mode driver will
be able to bring down the system.

Guest

Posted: Thu Jun 09, 2005 1:57 pm

Quote:

It's a driver issue.

Prove it.

It works fine with some video card/driver combinations. It fails with
the others. When one user in this thread had the crash (3rd or 4th post
- see screenshot) on startup Windows displayed a dialog listing the
name of the DRIVER that crashed.

Quote:

No different than the recent bug where...
tar bringing down a Linux system?

Never had that happen.

Then you didn't have a Linux install that used the faulty driver. Just
like my XP install doesn't have a video driver that crashes with this
rogue image.

robert

Posted: Thu Jun 09, 2005 2:01 pm

Guest

Tim Smith <reply_in_group@mouse-potato.com> wrote:

Quote:

In article <7-ednYAA2oJ98zXfRVn-sg@ctc.net>, robert wrote:
The odds that everyone who has had this problem is using the same video driver
are probably pretty long. Did you test with multiple src's as suggested

How so? There are only a handful of video drivers nowadays.

Here's some of the nvidia, ati and matrox drivers available for
XP/2000. It's tad more than a "handful":

http://www.3dchipset.com/drivers/nvidia/official/nt5/index.php

71.89 Forceware
Apr 14, 2005

19.40Mb

Download Here
67.66 Forceware
Jan 18, 2005

13.66Mb

Download Here
61.76 Forceware
Jul 20, 2004

20.60Mb

Download Here
56.72 Forceware
Apr 01, 2004

17.70Mb

Download Here
53.03 Forceware
Dec 09, 2003

12.96Mb

Download Here
52.16 Forceware
Oct 23, 2003

18.50Mb

Download Here
45.33 Detonator
Sep 09, 2003

9.50Mb

Download Here

http://www.3dchipset.com/drivers/nvidia/beta/nt5/index.php

77.13 Forceware
Jun 04, 2005

16.71Mb

Download Here
76.50 Forceware
Apr 09, 2005

19.44Mb

Download Here
76.45 Forceware
May 04, 2005

23.18Mb

Download Here
76.44 Forceware
Apr 07, 2005

23.17Mb

Download Here
76.41 Forceware
Mar 25, 2005

19.06Mb

Download Here
76.10 Forceware
Mar 10, 2005

23.25Mb

Download Here
75.90 Forceware
Feb 22, 2005

23.38Mb

Download Here
73.00 Forceware
May 12, 2005

19.20Mb

Download Here
71.90 Forceware
Mar 03, 2005

22.99Mb

Download Here
71.89 Forceware
Apr 09, 2005

17.36Mb

Download Here
71.84 Forceware
Feb 26, 2005

23.24Mb

Download Here
71.81 Forceware
Feb 15, 2005

22.98Mb

Download Here
71.80 Forceware
Jan 31, 2005

18.07Mb

Download Here
71.50 Forceware
Jan 15, 2005

22.90Mb

Download Here
71.40 Forceware
Jan 14, 2005

23.02Mb

Download Here
71.25 Forceware
Jan 14, 2005

22.91Mb

Download Here
71.24 Forceware
Dec 29, 2004

27.71Mb

Download Here
71.20 Forceware
Dec 13, 2004

22.92Mb

Download Here
70.90 Forceware
Dec 07, 2004

27.59Mb

Download Here
70.41 Forceware
Oct 25, 2004

18.48Mb

Download Here

http://www.3dchipset.com/drivers/nvidia/xtreme_g/nt5/index.php

76.50 HD Xtreme G
May 07, 2005

17.92Mb

Download Here
75.90a Xtreme G
Feb 27, 2005

21.51Mb

Download Here
71.24 Alpha Xtreme G
Jan 05, 2005

27.39Mb

Download Here
71.20c NV3x Xtreme G
Dec 27, 2004

23.32Mb

Download Here
71.20b NV4x Xtreme G
Dec 27, 2004

23.32Mb

Download Here
71.20 Xtreme G
Dec 16, 2004

23.32Mb

Download Here
70.90 Xtreme G
Dec 07, 2004

28.01Mb

Download Here

http://www.3dchipset.com/drivers/nvidia/firecat/nt5/index.php

Firecat 61.40 Forceware
Jun 08, 2004

8.93Mb

Download Here
Firecat 56.55 Forceware
Feb 16, 2004

7.50Mb

Download Here
Firecat 53.30 Forceware
Dec 17, 2003

5.12Mb

Download Here
Firecat 53.04 Forceware
Dec 13, 2003

7.46Mb

Download Here
Firecat 53.03 Forceware
Nov 19, 2003

7.50Mb

Download Here

http://www.3dchipset.com/drivers/ati/official/index.php

Catalyst 5.6
Jun 09, 2005

24.99Mb

Download Here
Catalyst 5.5
May 17, 2005

24.90Mb

Download Here
Catalyst 5.4
Apr 07, 2005

24.60Mb

Download Here
Catalyst 5.3
Mar 09, 2005

23.20Mb

Download Here
Catalyst 5.2
Feb 09, 2005

23.50Mb

Download Here
Catalyst 5.1
Jan 17, 2005

22.80Mb

Download Here

http://www.3dchipset.com/drivers/ati/omega/nt5/index.php

Omega Drivers 2.6.12
Apr 02, 2005

16.82Mb

Download Here
Omega Drivers 2.5.97
Dec 04, 2004

18.50Mb

Download Here
Omega Drivers 2.5.90
Oct 02, 2004

17.68Mb

Download Here
Omega Drivers 2.5.58
Jul 17, 2004

19.47Mb

Download Here
Omega Drivers 2.5.51
Jun 17, 2004

22.47Mb

Download Here

http://www.matrox.com/mga/support/drivers/latest/home.cfm

5.93.009
26oct04

5.82.018
26feb02
(final)

DFS

Posted: Thu Jun 09, 2005 2:07 pm

Guest

Quote:

Peter Köhlmann wrote:
Try this with XP (all patched)

HTML
BODY
IMG SRC="./sweetydead.jpg" width="9999999" height="9999999"
/BODY
/HTML

The image does not need to be that large, a very small one is
sufficient Just the /declaration/ of the size is enough to trigger
the bug

Save your data before. The machine will BSOD

Total lie.

I'm sitting here looking at that link
http://telon.dynalias.net/crashxp/crashxp.html in IE 6. In fact, I opened
up 5 instances of it in IE and 5 instances of it in Firefox, all at once.
And not a thing happened: no system slowdown, no excessive CPU usage,
nothing.

Windows Server 2003, 1gig RAM, ATI 9600 video card

Now had you said "Some web browsers running on some versions of Windows
running some display drivers can't handle some large images" you *might*
have been right.

But what else to expect from you but a lie.

Guest

Posted: Thu Jun 09, 2005 2:08 pm

Quote:

URL, please. With a date.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=145563

It's recent enough - from 2005.

So what's the "big-story" here? That kernel mode drivers are capable of
taking down a OS. Hardly news is it?

Page 3 of 10 Goto page Previous 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 Next

All times are GMT - 5 Hours The time now is Mon Nov 30, 2009 3:19 am Linux Forum Index