Skip to contentSkip to conversation

pahole BRAIN FART ALERT on F12

DISCUSSION Mike Snitzer 2 participants 7 messages Nov 23, 2009 – Nov 24, 2009 6 messages not in the archive

lore
Share & tools
lore
#0 lore raw

Hi,

I just found out about pahole today. I'm looking to audit lvm2's data structures and pahole is perfect for helping me do so.

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */
        int                        calories;             /*     0     4 */
        short int                  price;                /*     0     2 */
        int                        barcode[4];           /*     0    16 */
        /* size: 48, cachelines: 1, members: 6 */
        /* padding: 32 */
        /* last cacheline: 48 bytes */

        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */

};

When I compare the above to the output listed in your paper it is clear pahole is pretty unhappy on F12. If I do the same on F11 (x86_64) all works fine.

Should I open a fedora (F12) bug?

Mike
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
In reply to a message not in the archive

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 25 lines

Hi,

I just found out about pahole today. I'm looking to audit lvm2's data structures and pahole is perfect for helping me do so.

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */
        int                        calories;             /*     0     4 */
        short int                  price;                /*     0     2 */
        int                        barcode[4];           /*     0    16 */
        /* size: 48, cachelines: 1, members: 6 */
        /* padding: 32 */
        /* last cacheline: 48 bytes */

        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
Quoted 7 lines

};

When I compare the above to the output listed in your paper it is clear pahole is pretty unhappy on F12. If I do the same on F11 (x86_64) all works fine.

Should I open a fedora (F12) bug?

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

Some milliseconds ago I was almost going to suggest you using the git version, but then I realized this is too strange to seem like something I could have fixed in the git repository :-\

That also should be another reason for me to do the f11->f12 step :-)

Thanks!

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#4 lore raw
In reply to a message not in the archive
Quoted 38 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 25 lines

Hi,

I just found out about pahole today. I'm looking to audit lvm2's data structures and pahole is perfect for helping me do so.

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */
        int                        calories;             /*     0     4 */
        short int                  price;                /*     0     2 */
        int                        barcode[4];           /*     0    16 */
        /* size: 48, cachelines: 1, members: 6 */
        /* padding: 32 */
        /* last cacheline: 48 bytes */

        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
Quoted 7 lines

};

When I compare the above to the output listed in your paper it is clear pahole is pretty unhappy on F12. If I do the same on F11 (x86_64) all works fine.

Should I open a fedora (F12) bug?

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

Quoted 5 lines

Some milliseconds ago I was almost going to suggest you using the git version, but then I realized this is too strange to seem like something I could have fixed in the git repository :-\

That also should be another reason for me to do the f11->f12 step :-)

Yes... embrace F12... :)

BTW, I did try the latest git version from:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git

It didn't work either.

Mike
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#6 lore raw
In reply to a message not in the archive

Em Mon, Nov 23, 2009 at 05:56:15PM -0500, Mike Snitzer escreveu:

Quoted 12 lines
Quoted 10 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 9 lines

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */

<SNIP>

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines
        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
};

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

BTW, I did try the latest git version from:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git
It didn't work either.

Can you try with the attached patch applied against the git repo version?

Best Regards,

- Arnaldo

dwarf_loader.c +17 -3
index 44605a8..3b58ecb 100644
@@ -242,9 +242,23 @@ static Dwarf_Off attr_offset(Dwarf_Die *die, const uint32_t name)
 	Dwarf_Attribute attr;
 	Dwarf_Block block;
 
-	if (dwarf_attr(die, name, &attr) != NULL &&
-	    dwarf_formblock(&attr, &block) == 0)
-		return dwarf_expr(block.data, block.length);
+	if (dwarf_attr(die, name, &attr) == NULL)
+		return 0;
+
+	switch (dwarf_whatform(&attr)) {
+	case DW_FORM_data1:
+	case DW_FORM_data2:
+	case DW_FORM_sdata:
+	case DW_FORM_udata: {
+		Dwarf_Word value;
+		if (dwarf_formudata(&attr, &value) == 0)
+			return value;
+		break;
+	}
+	default:
+		if (dwarf_formblock(&attr, &block) == 0)
+			return dwarf_expr(block.data, block.length);
+	}
 
 	return 0;
 }
#8 lore raw
In reply to a message not in the archive
Quoted 20 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:56:15PM -0500, Mike Snitzer escreveu:

Quoted 12 lines
Quoted 10 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 9 lines

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */

<SNIP>

Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
};

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

Quoted 6 lines
Quoted 3 lines

BTW, I did try the latest git version from:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git
It didn't work either.

Can you try with the attached patch applied against the git repo version?

Works great, thanks!

Would you still like me to file an F12 bug?

Mike
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#10 lore raw
In reply to a message not in the archive

Em Tue, Nov 24, 2009 at 08:28:31AM -0500, Mike Snitzer escreveu:

Quoted 26 lines
Quoted 24 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:56:15PM -0500, Mike Snitzer escreveu:

Quoted 12 lines
Quoted 10 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 9 lines

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */

<SNIP>

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines
        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
};

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

BTW, I did try the latest git version from:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git
It didn't work either.

Quoted 6 lines
Quoted 2 lines

Can you try with the attached patch applied against the git repo version?

Works great, thanks!

Would you still like me to file an F12 bug?

Please, as I'll have to go the process for updating the package... :-)

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#12 lore raw
In reply to a message not in the archive
Quoted 27 lines · Arnaldo de Melo

Em Tue, Nov 24, 2009 at 08:28:31AM -0500, Mike Snitzer escreveu:

Quoted 26 lines
Quoted 24 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:56:15PM -0500, Mike Snitzer escreveu:

Quoted 12 lines
Quoted 10 lines · Arnaldo de Melo

Em Mon, Nov 23, 2009 at 05:17:03PM -0500, Mike Snitzer escreveu:

Quoted 9 lines

Unfortunately, when I run pahole against lvm2 I'm getting many BRAIN FART ALERTs on F12 x86_64 (dwarves-1.7-5.x86_64). I had a look at your OLS 2007 "7 dwarves" paper and figured I'd try your minimalist swiss_cheese example, this is what I get:

$ pahole swiss_cheese
struct cheese {
        char                       name[17];             /*     0    17 */
        short int                  age;                  /*     0     2 */
        char                       type;                 /*     0     1 */

<SNIP>

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines
        /* BRAIN FART ALERT! 48 != 17 + 0(holes), diff = 31 */
};

You should, I think this is related to some recent elfutils changes that Mark Wieelard warned me about it but I hadn't time yet to act upon, will do so tomorrow, thanks for the report!

BTW, I did try the latest git version from:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git
It didn't work either.

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

Can you try with the attached patch applied against the git repo version?

Works great, thanks!

Would you still like me to file an F12 bug?

Please, as I'll have to go the process for updating the package... :-)

Done, please see RH bz # 540955
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html