Skip to contentSkip to conversation

of dump command

PATCH New Jerry Van Baren 3 participants 4 messages Oct 30, 2006 – Oct 31, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 of dump command

See attached patch or get from the git repo:
<http://www.cideas.us/cgi-bin/gitweb.cgi?p=u-boot/u-boot-of-cmd.git;a=summary>

This should apply cleanly to the head of the denx.de git repository and works in conjunction with Jon Loeliger and others OF improvements.

Note that I've extended the command enable/disable mechanism by simply adding a '1' after CONFIG/CFG, creating CONFIG1_COMMANDS and CFG1_CMD_OF. This seemed to be the simplest way of extending the command selection.

Best regards,
gvb
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 061028-of-cmd.patch
Url: http://lists.denx.de/pipermail/u-boot/attachments/20061030/d894e4b6/attachment.txt

#1 lore raw

PATCH 1/1 of dump command

Quoted 10 lines · Jerry Van Baren

commit e57c182cdbd382e07338cfb395d8f4b98772913f

Author: Jerry Van Baren <[email protected]>
Date:   Sat Oct 28 12:17:52 2006 -0400

Add a "of" command to dump the flattened open firmware tree.

Enhanced the ft_build tree dump to allow dumping from a variable instead of always dumping the whole tree. Also fixed some leading zero formatting and printing 8 byte variables when 64 bit vsprintf is not supported.

Quoted 5 lines

Enhance vsprintf to handle "ll" long long specifier (it already supported the deprecated "q" specifier).

Expanded the command configuration into another long long flag word since the command flag is (almost) out of bits.

These should probably be split up into separate patches. One for the vsprintf 'll' support; one for the new command and one for the MPC8260ADS config changes.

Quoted 18 lines
common/Makefile +1 -1
index 07ddc95..4d883f6 100644
@@ -37,7 +37,7 @@ COBJS = main.o ACEX1K.o altera.o bedbug.
          cmd_i2c.o cmd_ide.o cmd_immap.o cmd_itest.o cmd_jffs2.o \
          cmd_load.o cmd_log.o \
          cmd_mem.o cmd_mii.o cmd_misc.o cmd_mmc.o \
-         cmd_nand.o cmd_net.o cmd_nvedit.o \
+         cmd_nand.o cmd_net.o cmd_nvedit.o cmd_of.o \
          cmd_pci.o cmd_pcmcia.o cmd_portio.o \
          cmd_reginfo.o cmd_reiser.o cmd_scsi.o cmd_spi.o cmd_universe.o \
          cmd_usb.o cmd_vfd.o \
common/cmd_bootm.c +0 -0
index 3091a58..58f50bc 100644

common/cmd_bootm.c #else /* CONFIG_OF_FLAT_TREE */ · line 953

Quoted 5 lines
common/cmd_bootm.c +1 -1
@@ -950,7 +950,7 @@ #else       /* CONFIG_OF_FLAT_TREE */
        }
 
        ft_setup(of_flat_tree, kbd, initrd_start, initrd_end);
-       /* ft_dump_blob(of_flat_tree); */
+       /* ft_dump_blob(of_flat_tree, "/"); */

This is dead code; just remove it; it's been replaced by a calls to ft_dump_blob() in ft_setup().

common/cmd_bootm.c #endif · line 971

Quoted 9 lines
common/cmd_bootm.c +1 -1
 #if defined(CFG_INIT_RAM_LOCK) && !defined(CONFIG_E500)
        unlock_ram_in_cache();
@@ -968,7 +968,7 @@ #endif
                        cmd_start, cmd_end);
        else {
                ft_setup(of_flat_tree, kbd, initrd_start, initrd_end);
-               /* ft_dump_blob(of_flat_tree); */
+               /* ft_dump_blob(of_flat_tree, "/"); */

ditto

Quoted 30 lines
common/cmd_bootm.c +0 -0
                (*kernel) ((bd_t *)of_flat_tree, (ulong)kernel, 0, 0, 0);
        }
 #endif /* CONFIG_OF_FLAT_TREE */
common/cmd_of.c new file +21 -0
new file mode 100644
index 0000000..0f7973f
@@ -0,0 +1,58 @@
+/*
+ * (C) Copyright 2006
+ * Gerald Van Baren, Custom IDEAS, vanbaren at cideas.com.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA

common/cmd_of.c line 31

Quoted 10 lines
common/cmd_of.c new file +10 -0
+ */
+
+/*
+ * Misc functions
+ */
+#include <common.h>
+#include <command.h>
+#include <ft_build.h>
+
+#if (CONFIG1_COMMANDS & CFG1_CMD_OF)

This is not a lot of code; and it's pretty useful stuff; maybe just make it automatically built when CONFIG_OF_FLAT_TREE is defined? (Rather than adding a new CONFIG_COMMANDS macro at the moment)

If a new CFG_CMD_ macro set is needed; I wouldn't do it now. There is still 1 bit left. :) If it needs to be changed, it might make sense to audit the full list and see if there is anything that should be removed.

Wolfgang needs to weigh in on this one.

Quoted 14 lines
common/cmd_of.c new file +14 -0
+
+int do_of (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
+{
+       void    *p = (void *)simple_strtoul(argv[1], NULL, 16);
+
+       if (argc < 2) {
+               printf ("Usage:\n%s\n", cmdtp->usage);
+               return 1;
+       } else if (argc == 2) {
+               ft_dump_blob(p, "/");
+       } else {
+               ft_dump_blob(p, argv[2]);
+       }
+

common/cmd_of.c line 55

Quoted 10 lines
common/cmd_of.c new file +10 -0
+       return 0;
+}
+
+
+U_BOOT_CMD(
+       of,     3,      0,      do_of,
+       "of      - Open Firmware utility commands\n",
+       "of <addr>        - Dump the address as an OF tree\n"
+       "of <addr> <prop> - Dump the given property\n"
+);

I think 'of' is a bit terse. Maybe "ofinfo", "ofdump", or "ftinfo"?

Quoted 30 lines
common/cmd_of.c new file +3 -0
+
+#endif /* CFG1_CMD_OF */
+
common/ft_build.c +11 -2
index 980e40f..a752b5e 100644
@@ -233,10 +233,19 @@ static void print_data(const void *data,
                printf(" = <%04x>", be16_to_cpu(*(u16 *) data) & 0xffff);
                break;
        case 4:         /* word */
-               printf(" = <%x>", be32_to_cpu(*(u32 *) data) & 0xffffffffU);
+               printf(" = <%08x>", be32_to_cpu(*(u32 *) data) & 0xffffffffU);
                break;
        case 8:         /* double-word */
-               printf(" = <%qx>", be64_to_cpu(*(uint64_t *) data));
+               /*
+                * If 64 bit printing is supported, use it.
+                */
+#ifdef CFG_64BIT_VSPRINTF
+               printf(" = <%016llx>", be64_to_cpu(*(uint64_t *) data));
+#else
+               printf(" = <%08x", be32_to_cpu(*(u32 *) data) & 0xffffffffU);
+               data += sizeof(u32);
+               printf(" %08x>", be32_to_cpu(*(u32 *) data) & 0xffffffffU);
+#endif
                break;
        default:                /* anything else... hexdump */
                printf(" = [");

common/ft_build.c static void print_data(const void *data, · line 263

Quoted 8 lines
common/ft_build.c +4 -1
@@ -248,7 +257,12 @@ static void print_data(const void *data,
        }
 }
 
-void ft_dump_blob(const void *bphp)
+/*
+ * Used by ft_dump_blob() and ft_get_prop() to build up property names
+ */
+static char path[256], prop[256];

Is moving these out of ft_get_prop truly necessary? (I haven't dug into the full context, but seeing this makes me scared)

Quoted 69 lines
common/ft_build.c +28 -5
+
+void ft_dump_blob(const void *bphp, const char *propname)
 {
        const struct boot_param_header *bph = bphp;
        const uint64_t *p_rsvmap = (const uint64_t *)
@@ -260,7 +274,9 @@ void ft_dump_blob(const void *bphp)
        u32 tag;
        const u32 *p;
        const char *s, *t;
+       char *ss;
        int depth, sz, shift;
+       int found;
        int i;
        uint64_t addr, size;
 
@@ -271,6 +287,8 @@ void ft_dump_blob(const void *bphp)
 
        depth = 0;
        shift = 4;
+       found = 0;
+       strcpy(path, "/");
 
        for (i = 0;; i++) {
                addr = be64_to_cpu(p_rsvmap[i * 2]);
@@ -290,7 +308,17 @@ void ft_dump_blob(const void *bphp)
                        s = (const char *)p;
                        p = (u32 *) _ALIGN((unsigned long)p + strlen(s) + 1, 4);
 
-                       printf("%*s%s {\n", depth * shift, "", s);
+                       strcat(path, s);
+                       if (found || (strcmp(path, propname) == 0)) {
+                               found++;
+                               if (depth == 0)
+                                       printf("/ {\n");
+                               else
+                                       printf("%*s%s {\n", depth * shift, "", s);
+                       }
+                       /* The root path is already there as "/" */
+                       if(depth != 0)
+                               strcat(path, "/");
 
                        depth++;
                        continue;
@@ -299,7 +327,16 @@ void ft_dump_blob(const void *bphp)
                if (tag == OF_DT_END_NODE) {
                        depth--;
 
-                       printf("%*s};\n", depth * shift, "");
+                       path[strlen(path) - 1] = '\0';
+                       ss = strrchr(path, '/');
+                       if (ss != NULL)
+                               ss[1] = '\0';
+
+                       if(found) {
+                               printf("%*s};\n", depth * shift, "");
+                               if (found-- == 0)
+                                       return;         /* request done */
+                       }
                        continue;
                }
 
@@ -317,9 +354,15 @@ void ft_dump_blob(const void *bphp)
                s = (const char *)p_strings + be32_to_cpu(*p++);
                t = (const char *)p;
                p = (const u32 *)_ALIGN((unsigned long)p + sz, 4);
-               printf("%*s%s", depth * shift, "", s);
-               print_data(t, sz);
-               printf(";\n");
+

common/ft_build.c void ft_dump_blob(const void *bphp) · line 367

Quoted 10 lines
common/ft_build.c +8 -0
+               strcpy(prop, path);
+               strcat(prop, s);
+
+               if(found || (strcmp(prop, propname) == 0)) {
+                       printf("%*s%s", depth * shift, "", s);
+                       print_data(t, sz);
+                       printf(";\n");
+               }
        }
 }

Hmm, I'm not all that thrilled with this approach; but I understand that some of it is driven by the current design of the ft_dump_blob code. I'd rather have a block at the top which 'seeks' to the correct point in the device tree and sets the depth value to zero. Then start printing the tree until depth gets back to zero.

Quoted 32 lines
common/ft_build.c +2 -3
@@ -349,7 +392,6 @@ void *ft_get_prop(void *bphp, const char
        char *s, *t;
        char *ss;
        int sz;
-       static char path[256], prop[256];
 
        path[0] = '\0';
 
@@ -478,7 +520,7 @@ #endif
 
 #ifdef DEBUG
        printf ("recieved oftree\n");
-       ft_dump_blob(blob);
+       ft_dump_blob(blob, "/");
 #endif
 
        ft_init_cxt(&cxt, blob);
@@ -585,7 +627,7 @@ #endif
 
 #ifdef DEBUG
        printf("final OF-tree\n");
-       ft_dump_blob(blob);
+       ft_dump_blob(blob, "/");
 #endif
 }
 #endif
include/cmd_confdefs.h +0 -0
index cf36583..52b873d 100644

include/cmd_confdefs.h #define CFG_CMD_UNIVERSE 0x0800000000000 · line 100

Quoted 7 lines
include/cmd_confdefs.h +4 -0
@@ -94,8 +94,13 @@ #define CFG_CMD_UNIVERSE 0x0800000000000
 #define CFG_CMD_EXT2   0x1000000000000000ULL   /* EXT2 Support                 */
 #define CFG_CMD_SNTP   0x2000000000000000ULL   /* SNTP support                 */
 #define CFG_CMD_DISPLAY        0x4000000000000000ULL   /* Display support              */
+/*
+ * Start a second level of commands...
+ */
+#define CFG1_CMD_OF    0x0000000000000001ULL   /* Open Firmware dump support   */

There's still one more bit left in CFG_CMD_ALL; just use it! :)

Or eliminate CFG_CMD_OF entirely.

Quoted 18 lines
include/configs/MPC8260ADS.h +3 -0
index 6195bca..55c6ff9 100644
@@ -237,6 +237,7 @@ #define CONFIG_COMMANDS             (CFG_CMD_ALL &
                                 CFG_CMD_I2C    | \
                                 CFG_CMD_PCI    | \
                                 CFG_EXCLUDE    ) )
+#define CONFIG1_COMMANDS       (CFG1_CMD_OF)
 #else
 #define CONFIG_COMMANDS                (CFG_CMD_ALL & ~( \
                                 CMD_CFG_PCI    | \
@@ -506,4 +507,15 @@ #if CONFIG_ADSTYPE == CFG_8272ADS
 #define CONFIG_HAS_ETH1
 #endif
 
+/* pass open firmware flat tree */
+#define CONFIG_OF_FLAT_TREE     1

include/configs/MPC8260ADS.h #if CONFIG_ADSTYPE == CFG_8272ADS · line 521

Quoted 10 lines
include/configs/MPC8260ADS.h +9 -0
+
+/* maximum size of the flat tree (8K) */
+#define OF_FLAT_TREE_MAX_SIZE   8192
+
+#define OF_CPU                  "PowerPC,MPC8272 at 0"
+#define OF_TBCLK                (CONFIG_8260_CLKIN / 4)
+#define CONFIG_OF_HAS_BD_T      1
+#define CONFIG_OF_HAS_UBOOT_ENV 1
+
 #endif /* __CONFIG_H */

All of this should be in a separate patch

Quoted 18 lines
include/ft_build.h +1 -1
index 89c689c..5530122 100644
@@ -58,7 +58,7 @@ void ft_add_rsvmap(struct ft_cxt *cxt, u
 
 void ft_setup(void *blob, bd_t * bd, ulong initrd_start, ulong initrd_end);
 
-void ft_dump_blob(const void *bphp);
+void ft_dump_blob(const void *bphp, const char *propname);
 void ft_merge_blob(struct ft_cxt *cxt, void *blob);
 void *ft_get_prop(void *bphp, const char *propname, int *szp);
 
lib_generic/vsprintf.c +0 -0
index 2740f2e..034c619 100644

lib_generic/vsprintf.c #endif · line 265

Quoted 10 lines
lib_generic/vsprintf.c +7 -0
@@ -256,6 +256,13 @@ #endif
                if (*fmt == 'h' || *fmt == 'l' || *fmt == 'q') {
                        qualifier = *fmt;
                        ++fmt;
+#ifdef CFG_64BIT_VSPRINTF
+                       /* parse ll (64 bit) and change to 'q' */
+                       if (*fmt == 'l') {
+                               qualifier = 'q';
+                               ++fmt;
+                       }
+#endif

ditto

Signature · 4 lines
Grant Likely, B.Sc. P.Eng.
Secret Lab Technologies Ltd.
grant.likely at secretlab.ca
(403) 399-0195
#2 lore raw

PATCH 1/1 of dump command

Dear Grant,

thanks a lot for your thorough comments.

Quoted 4 lines · In message <[email protected]> you…

These should probably be split up into separate patches. One for the vsprintf 'll' support; one for the new command and one for the MPC8260ADS config changes.

Agreed. Eventually these are separate commits in Jerry's repo anyway. Jerry?

Quoted 3 lines

This is not a lot of code; and it's pretty useful stuff; maybe just make it automatically built when CONFIG_OF_FLAT_TREE is defined? (Rather than adding a new CONFIG_COMMANDS macro at the moment)

Agreed.

Quoted 6 lines

If a new CFG_CMD_ macro set is needed; I wouldn't do it now. There is still 1 bit left. :) If it needs to be changed, it might make sense to audit the full list and see if there is anything that should be removed.

Wolfgang needs to weigh in on this one.

I think we need a major cleanup of the config system, so I would like to touch as few things in this area as possible. Here, I agree with Grant that this should be just compiled in when OFT support is enabled. Probably everybody will want this.

Quoted 8 lines

Review

Quoted 6 lines
(quoted diff) +6 -0
+U_BOOT_CMD(
+       of,     3,      0,      do_of,
+       "of      - Open Firmware utility commands\n",
+       "of <addr>        - Dump the address as an OF tree\n"
+       "of <addr> <prop> - Dump the given property\n"
+);

I think 'of' is a bit terse. Maybe "ofinfo", "ofdump", or "ftinfo"?

Additionally, I think that "of" [= Open Firmware] is wrong. It should be "oft" [Open Firmware Tree] at least. The command should IMHO be named "oftdump", but "ftdump" [FT = Flat Tree] would be ok, too.

Quoted 6 lines
Quoted 3 lines

+ * Used by ft_dump_blob() and ft_get_prop() to build up property names
+ */
+static char path[256], prop[256];

Is moving these out of ft_get_prop truly necessary? (I haven't dug into the full context, but seeing this makes me scared)

Me too :-)

Quoted 3 lines

There's still one more bit left in CFG_CMD_ALL; just use it! :)

Or eliminate CFG_CMD_OF entirely.

That's what I'd suggest.

Thanks again, to both of you.

Best regards,

Wolfgang Denk

Signature · 2 lines
Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Quoted 1 line

Is there a way to determine Yesterday's date using Unix utilities?

echo "what is yesterday's date?" | /bin/mail root -- Randal L. Schwartz in <[email protected]>

#3 lore raw

PATCH 1/1 of dump command

Wow, you guys are _tough_! I'm glad you didn't grade me for CompSci 101
:-)

Three patches will follow
* vsprintf "ll" extension (pretty trivial)
* oftdump command (da meat)
* PQ2FADS changes to use the flattened OF tree (pretty trivial)

More comments embedded...

Quoted 11 lines · Wolfgang Denk

Dear Grant,

thanks a lot for your thorough comments.

Quoted 3 lines · In message <[email protected]> you…

These should probably be split up into separate patches. One for the vsprintf 'll' support; one for the new command and one for the MPC8260ADS config changes.

Agreed. Eventually these are separate commits in Jerry's repo anyway. Jerry?

Done, patch will be submitted.

Quoted 5 lines
Quoted 3 lines

This is not a lot of code; and it's pretty useful stuff; maybe just make it automatically built when CONFIG_OF_FLAT_TREE is defined? (Rather than adding a new CONFIG_COMMANDS macro at the moment)

Agreed.

Done, patch will be submitted.

Quoted 11 lines
Quoted 6 lines

If a new CFG_CMD_ macro set is needed; I wouldn't do it now. There is still 1 bit left. :) If it needs to be changed, it might make sense to audit the full list and see if there is anything that should be removed.

Wolfgang needs to weigh in on this one.

I think we need a major cleanup of the config system, so I would like to touch as few things in this area as possible. Here, I agree with Grant that this should be just compiled in when OFT support is enabled. Probably everybody will want this.

CONFIG1_* and CFG1_CMD_* have been removed.

Quoted 11 lines
Quoted 7 lines

Review

Quoted 6 lines
(quoted diff) +6 -0
+U_BOOT_CMD(
+       of,     3,      0,      do_of,
+       "of      - Open Firmware utility commands\n",
+       "of <addr>        - Dump the address as an OF tree\n"
+       "of <addr> <prop> - Dump the given property\n"
+);

I think 'of' is a bit terse. Maybe "ofinfo", "ofdump", or "ftinfo"?

Additionally, I think that "of" [= Open Firmware] is wrong. It should be "oft" [Open Firmware Tree] at least. The command should IMHO be named "oftdump", but "ftdump" [FT = Flat Tree] would be ok, too.

OK, I renamed it "oftdump" (but, thanks to the beauty of Wolfgang's minimum unique characters parsing, "of" still works ;-).

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

+ * Used by ft_dump_blob() and ft_get_prop() to build up property names
+ */
+static char path[256], prop[256];

Is moving these out of ft_get_prop truly necessary? (I haven't dug into the full context, but seeing this makes me scared)

Me too :-)

Yes, this is ugly. The existing ft_get_prop() had the two declarations as statics within the function declaration and I simply moved them out so they could be re-used by ft_dump_blob(). Note that they remain static, but they are available to ft_dump_blob() as well as ft_get_prop() now. One alternative is to duplicate the static declarations in both routines, but that seems extremely silly.

(I'm also not wild about the fact that the buffer lengths are fixed and unchecked, so it is a buffer overflow vulnerability - I figure (a) I just used existing code and (b) simply having access to u-boot is more of a security issue than buffer overflows in u-boot.)

The problem is that the flattened OF tree only has the incremental path/property names rather than the full name on each property (I believe this was a version 0x10 optimization). The easiest way to match the path/property is to build up (as you nest in) and back up (as you unnest) the path and then do a string compare of the path/property built-up string against the desired string. This is what the existing ft_get_prop() did and what I extended ft_dump_blob() to do too.

I looked at least twice at doing a more clever matching and threw my changes away because I ended up going down a blind alley. Keeping track of the nesting and unnesting while following the tree takes some non-trivial logic. I'm not saying that there _isn't_ a better way, but building up the string and doing a string compare is _much_ easier and quite likely takes less code than being clever.

Quoted 5 lines
Quoted 3 lines

There's still one more bit left in CFG_CMD_ALL; just use it! :)

Or eliminate CFG_CMD_OF entirely.

That's what I'd suggest.

Done.

Quoted 5 lines

Thanks again, to both of you.

Best regards,

Wolfgang Denk

Patches to follow...

Best regards,
gvb