[U-Boot-Users] [PATCH] mpc885ads: Don't define CONFIG_BZIP2.

bzip2 requires a significant chunk of malloc space, and there isn't enough room on mpc885ads (with only 8MB RAM) for both bzip2's malloc area and a downloaded image at 0x400000.
Signed-off-by: Scott Wood scottwood@freescale.com --- board/fads/fads.h | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/board/fads/fads.h b/board/fads/fads.h index c6f7ccd..2d0a6c0 100644 --- a/board/fads/fads.h +++ b/board/fads/fads.h @@ -71,7 +71,10 @@ #undef CONFIG_BOOTARGS
#undef CONFIG_WATCHDOG /* watchdog disabled */ + +#if !defined(CONFIG_MPC885ADS) #define CONFIG_BZIP2 /* include support for bzip2 compressed images */ +#endif
/* * New MPC86xADS and MPC885ADS provide two Ethernet connectivity options:

In message 20070815204646.GB17995@ld0162-tx32.am.freescale.net you wrote:
bzip2 requires a significant chunk of malloc space, and there isn't enough room on mpc885ads (with only 8MB RAM) for both bzip2's malloc area and a downloaded image at 0x400000.
So download the image to 0x200000 ?
Best regards,
Wolfgang Denk

Wolfgang Denk wrote:
In message 20070815204646.GB17995@ld0162-tx32.am.freescale.net you wrote:
bzip2 requires a significant chunk of malloc space, and there isn't enough room on mpc885ads (with only 8MB RAM) for both bzip2's malloc area and a downloaded image at 0x400000.
So download the image to 0x200000 ?
There's no device tree support yet, so the wrapper is being used, and that expects to be loaded at 0x400000. Even if that were changed, moving it would mean there's less room for the relocated kernel, which gets moved down to zero (there's currently no way to tell the wrapper that the kernel's already uncompressed, so don't bother moving it at all).
-Scott

Dear Scott,
in message 46C36D49.20301@freescale.com you wrote:
So download the image to 0x200000 ?
There's no device tree support yet, so the wrapper is being used, and that expects to be loaded at 0x400000. Even if that were changed,
Well, there are probably folks out there who continue to use 2.4 kernels (which on 8xx makes perfect sense to me).
moving it would mean there's less room for the relocated kernel, which gets moved down to zero (there's currently no way to tell the wrapper that the kernel's already uncompressed, so don't bother moving it at all).
Well, let's face it: if 0x200000 doesn't work as load address any more because your kernel is too big, then you won't get lucky with it at all on a 8 MB system.
I don't really care about the mpc885ads, but I feel you remove a feature which might be useful for some to overcome a restriction which does not really solve a problem, because you will just hang again in the next step.
But you decide...
Best regards,
Wolfgang Denk

Wolfgang Denk wrote:
in message 46C36D49.20301@freescale.com you wrote:
There's no device tree support yet, so the wrapper is being used, and that expects to be loaded at 0x400000. Even if that were changed,
Well, there are probably folks out there who continue to use 2.4 kernels (which on 8xx makes perfect sense to me).
IIRC, 2.4 also had 0x400000 as the default load address, though maybe those kernels were small enough that they managed to fit.
Well, let's face it: if 0x200000 doesn't work as load address any more because your kernel is too big, then you won't get lucky with it at all on a 8 MB system.
The kernel I tested on the board goes to 0x210d84, and that's without any initramfs... Sure, it's going to be tight, but it's not completely out of the question.
I don't really care about the mpc885ads, but I feel you remove a feature which might be useful for some to overcome a restriction which does not really solve a problem, because you will just hang again in the next step.
It solves the problem of not tying up over 1/4 of the board's RAM for what AFAICT is a not-very-frequently used feature (Linux doesn't use it, at least). It solves the problem of not being able to load a current Linux kernel on this board. The user can always re-enable bzip2 support if they really need it.
-Scott

In message 46C37B3C.7070805@freescale.com you wrote:
I don't really care about the mpc885ads, but I feel you remove a feature which might be useful for some to overcome a restriction which does not really solve a problem, because you will just hang again in the next step.
It solves the problem of not tying up over 1/4 of the board's RAM for what AFAICT is a not-very-frequently used feature (Linux doesn't use it, at least). It solves the problem of not being able to load a current Linux kernel on this board. The user can always re-enable bzip2 support if they really need it.
OK, I give in...
Best regards,
Wolfgang Denk

In message 20070815204646.GB17995@ld0162-tx32.am.freescale.net you wrote:
bzip2 requires a significant chunk of malloc space, and there isn't enough room on mpc885ads (with only 8MB RAM) for both bzip2's malloc area and a downloaded image at 0x400000.
Signed-off-by: Scott Wood scottwood@freescale.com
Applied. Thanks.
Best regards,
Wolfgang Denk
participants (2)
-
Scott Wood
-
Wolfgang Denk