Skip to contentSkip to conversation

ramdisk: u-boot no longer boots a ramdisk with Linux 2.6.27-rc4

U-Boot Heiko Schocher 2 participants 6 messages Sep 5, 2008

lore
Share & tools
lore
#0 lore raw

Hello Kumar,

I actually trying to boot a Linux 2.6.27-rc4 Kernel with a Ramdisk and actual u-boot on a MPC82xx based board. And what should I say, it doesnt work anymore :-(

Reason is the following commit:

http://git.denx.de/?p=u-boot.git;a=commitdiff;h=2a1a2cb6e2b87ee550e6f27b647d23331dfd5e1b#patch3

After this commit, u-boot no longer deletes the mem reservation for the initrd in Flash, what stops Linux from booting.

So I did the following, but maybe there is a better way for it?

common/fdt_support.c +0 -2
index a7773ab..a69f001 100644
@@ -215,8 +215,6 @@ int fdt_chosen(void *fdt, ulong initrd_start, ulong initrd_end, int force)
 		}
 	}
 
-	fdt_initrd(fdt, initrd_start, initrd_end, force);
-
 #ifdef CONFIG_OF_STDOUT_VIA_ALIAS
 	path = fdt_getprop(fdt, nodeoffset, "linux,stdout-path", NULL);
 	if ((path == NULL) || force)

With this patch Linux boots fine again with the Ramdisk ...

Do you know, why fdt_chosen() is creating a dummy entry for the initrd, when later fdt_initrd() is again called with the right parameters?

I think, we can drop this!

bye,
Heiko

Signature · 2 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
#1 lore raw
Quoted 6 lines · Heiko Schocher

Hello Kumar,

I actually trying to boot a Linux 2.6.27-rc4 Kernel with a Ramdisk and actual u-boot on a MPC82xx based board. And what should I say, it doesnt work anymore :-(

Quoted 8 lines

Reason is the following commit:

http://git.denx.de/?p=u-boot.git;a=commitdiff;h=2a1a2cb6e2b87ee550e6f27b647d23331dfd5e1b#patch3

After this commit, u-boot no longer deletes the mem reservation for the initrd in Flash, what stops Linux from booting.

So I did the following, but maybe there is a better way for it?

I don't follow.. can you provide a bit more info about what's going on w/your mem reserves and what locations the ramdisk is at.

The code should have been identical to before.

- k

#2 lore raw

Hello Kumar,

Quoted 10 lines · Kumar Gala
Quoted 9 lines · Heiko Schocher

I actually trying to boot a Linux 2.6.27-rc4 Kernel with a Ramdisk and actual u-boot on a MPC82xx based board. And what should I say, it doesnt work anymore :-(

Reason is the following commit:

http://git.denx.de/?p=u-boot.git;a=commitdiff;h=2a1a2cb6e2b87ee550e6f27b647d23331dfd5e1b#patch3

Quoted 7 lines
Quoted 4 lines

After this commit, u-boot no longer deletes the mem reservation for the initrd in Flash, what stops Linux from booting.

So I did the following, but maybe there is a better way for it?

I don't follow.. can you provide a bit more info about what's going on w/your mem reserves and what locations the ramdisk is at.

OK, I try it ...

We start in lib_ppc/bootm.c (actual u-boot):

147 if (of_size) {
148 /* pass in dummy initrd info, we'll fix up later */
149 if (fdt_chosen(of_flat_tree, images->rd_start, images->rd_end, 0) < 0) {

This calls fdt_chosen with images->rd_start which are on my Hardware in Flash at 0xff21000

fdt_chosen calls

fdt_initrd(of_flat_tree, initrd_start, initrd_end, force);

with initrd = images->rd_start = 0xff210000

fdt initr ():
125 /*

 126          * Look for an existing entry and update it.  If we don't find
 127          * the entry, we will j be the next available slot.
 128          */

129 for (j = 0; j < total; j++) {
130 err = fdt_get_mem_rsv(fdt, j, &addr, &size);
131 if (addr == initrd_start) {
132 fdt_del_mem_rsv(fdt, j);
133 break;
134 }
135 }

didnt find this addres -> didnt call fdt_del_mem_rsv() This is OK here, because its the first call ...

fdt_initrd calls fdt_add_mem_rsv with this value
-> One dummy mem reserv for the 0xff210000 value

Ok lets go on, later in lib_ppc/bootm.c:

Here is called fdt_initrd again, now with the values from the initrd somewhere in RAM (which are the right ones) ...

181 #if defined(CONFIG_OF_LIBFDT)
182 /* fixup the initrd now that we know where it should be */
183 if ((of_flat_tree) && (initrd_start && initrd_end))
184 fdt_initrd(of_flat_tree, initrd_start, initrd_end, 1);
185 #endif

in fdt initr ():

125 /*

 126          * Look for an existing entry and update it.  If we don't find
 127          * the entry, we will j be the next available slot.
 128          */

129 for (j = 0; j < total; j++) {
130 err = fdt_get_mem_rsv(fdt, j, &addr, &size);
131 if (addr == initrd_start) {
132 fdt_del_mem_rsv(fdt, j);
133 break;
134 }
135 }

didnt find this addresses in RAM -> didnt call fdt_del_mem_rsv() for the dummy reservation (on my Hardware 0xff210000)...

This is NOT OK here!

Quoted 1 line

The code should have been identical to before.

Hmm... seems not so to me ...
... taking a look at your Patch:

Here your Patch for lib_ppc_bootm.c:

 #if defined(CONFIG_OF_LIBFDT)
        /* fixup the initrd now that we know where it should be */
-       if ((of_flat_tree) && (initrd_start && initrd_end)) {
-               uint64_t addr, size;
-               int  total = fdt_num_mem_rsv(of_flat_tree);
-               int  j;

-
- /* Look for the dummy entry and delete it */
- for (j = 0; j < total; j++) {
- fdt_get_mem_rsv(of_flat_tree, j, &addr, &size);
- if (addr == images->rd_start) {
- fdt_del_mem_rsv(of_flat_tree, j);

This looks for me that the old code explicitly searches for the dummy entry with images->rd_start!!

- break;
- }
- }
-
- ret = fdt_add_mem_rsv(of_flat_tree, initrd_start,
- initrd_end - initrd_start + 1);
- if (ret < 0) {
- printf("fdt_chosen: %s\n", fdt_strerror(ret));
- goto error;
- }
-
- do_fixup_by_path_u32(of_flat_tree, "/chosen",
- "linux,initrd-start", initrd_start, 0);
- do_fixup_by_path_u32(of_flat_tree, "/chosen",
- "linux,initrd-end", initrd_end, 0);
- }
+ if ((of_flat_tree) && (initrd_start && initrd_end))
+ fdt_initrd(of_flat_tree, initrd_start, initrd_end, 1);
^^^^^^^^^^^^
At this place (see explanation before) initrd_start has the value for the initrd in RAM! -> the search for this new address in fdt_initrd () is useless, even dangerous, if the initrd is not moved, to RAM this search is succesful and for this case the right mem reservation gets deleted ...
#endif

Nevertheless, the actual Code works with my patch fine, so again the question, why we need this dummy call in fdt_chosen for the initrd, when we call it later with the right values.

I think we can drop this useless dummy call from fdt_chosen. Has somebody any suggestions?

bye,
Heiko

Signature · 2 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
#3 lore raw
Quoted 2 lines

I think we can drop this useless dummy call from fdt_chosen. Has somebody any suggestions?

That breaks some of the commands. I see the problem and will post a fix for you to try.

- k

#4 lore raw

Hello Kumar,

Quoted 5 lines · Kumar Gala
Quoted 2 lines

I think we can drop this useless dummy call from fdt_chosen. Has somebody any suggestions?

That breaks some of the commands. I see the problem and will post a fix for you to try.

Which commands?

bye
Heiko

Signature · 2 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
#5 lore raw
Quoted 11 lines · Heiko Schocher

Hello Kumar,

Quoted 6 lines · Kumar Gala
Quoted 2 lines

I think we can drop this useless dummy call from fdt_chosen. Has somebody any suggestions?

That breaks some of the commands. I see the problem and will post a fix
for you to try.

Which commands?

fdt chosen

- k