drivers/cpu: add function to get CPU id/serial number #854

miri64 · 2014-03-04T13:10:24Z

See OlegHahm/thirdparty_cpu#19 for real world example.

LudwigKnuepfer · 2014-03-04T13:24:01Z

Where is the connection to #19 ?

LudwigKnuepfer · 2014-03-04T13:25:21Z

Ah I guess you used the wrong repo but the right number...

LudwigKnuepfer · 2014-03-04T13:26:06Z

Still it's circular reasoning - what should this be used for?

OlegHahm · 2014-03-04T13:43:18Z

I think it might be used to generate a host UID or EUI for the network interface.

miri64 · 2014-03-04T13:43:19Z

First: we discussed some time ago to use the cpu id as a (bad) source of randomness. Second: see #837

miri64 · 2014-03-04T13:49:16Z

(maybe @haukepetersen and @thomaseichinger could look over it, too, and give suggestions for other CPU implementations)

OlegHahm · 2014-03-04T14:01:35Z

ACK

LudwigKnuepfer · 2014-03-04T14:10:03Z

cpu/native/include/cpu.h

@@ -27,6 +27,9 @@
 /* TODO: choose better value? */
 #define F_CPU 1000000

+#define CPU_ID(id) id[0] = 1; id[1] = 2; id[2] = 3; id[3] = 4; \
+                   id[4] = 5; id[5] = 6; id[6] = 7; id[7] = 8; // TODO: find a source.


I'd go with the process ID (compare #700) for starters.

miri64 · 2014-03-04T14:34:04Z

Applied @LudwigOrtmann's and @kaspar030's comments.

kaspar030 · 2014-03-04T15:05:34Z

ACK for the core change.

mehlis · 2014-03-04T15:59:53Z

cpu/native/native_cpu.c

+    pid_t pid = getpid();
+
+    memset(id, 0, id_len);
+    memcpy(&id[0], &pid, id_len);


why a double memset?
why &id[0] same as id ?

If #854 (comment) is integrated it's needed.

miri64 · 2014-03-05T09:15:14Z

Added a commit with an alternative suggestion based on OlegHahm/thirdparty_cpu#21 (comment) please review.

LudwigKnuepfer · 2014-03-05T09:23:53Z

core/include/kernel.h

+ */
+typedef struct {
+    char cpu_id[CPU_ID_MAX_LEN];
+    int cpu_id_len;


What is cpu_id_len needed for? Can't the cpu set CPU_ID_MAX_LEN?

I mean - without cpu_id_len one could simply use sizeof(cpu_id_t)..

Then how do you propose to set CPU_ID_MAX_LEN to sizeof(pid_t) on native without including unistd.h and risking this way to come into conflict with our own posix port?

ah hm uhm eh... one could get nasty here and define it via make

I mean - have a test-program that determines this...

long can only be trusted to be >= int.

But at least for the *nix systems we support it seems to be the source for the typedef of pid_t

Alternatively we can just say cpu_id_len may allways be 4 in native, and if the pid does not supply enough bytes for the rest is set to 0xff or 0x00 or whatever.

The reason to trust pid_t to fit in long would be that:

blksize_t, pid_t, and ssize_t shall be signed integer types

The implementation shall support one or more programming environments in which the widths of blksize_t, pid_t, size_t, ssize_t, suseconds_t, and useconds_t are no greater than the width of type long.

(http://pubs.opengroup.org/onlinepubs/009695399/basedefs/sys/types.h.html)

And yes, I'd memset it before like @authmillenon said.

miri64 · 2014-03-06T17:17:47Z

Applied discussed changes and rebased. Also: updated #837 accordingly.

LudwigKnuepfer · 2014-03-07T10:15:23Z

ACK

miri64 · 2014-03-07T12:40:59Z

Needs second ACK because of core change (Re-ACK from @kaspar030?)

Kijewski · 2014-03-07T15:47:10Z

cpu/native/include/cpu.h

+extern void native_cpu_id(cpu_id_t *id);
+
+#undef GET_CPU_ID
+#define GET_CPU_ID(id)  native_cpu_id(&id)


Put parentheses around the macro argument.

miri64 · 2014-03-24T13:27:16Z

Rebased, squashed and applied suggestions.

thomaseichinger · 2014-05-05T18:16:49Z

needs rebase

miri64 · 2014-05-06T11:42:37Z

done

Kijewski · 2014-05-06T14:37:26Z

(still/again) needs rebase

LudwigKnuepfer · 2014-05-13T08:50:02Z

@authmillenon ping

OlegHahm · 2014-06-22T17:25:25Z

Please rebase.

miri64 · 2014-06-23T07:13:51Z

Done

miri64 · 2014-06-24T11:42:38Z

Uses functionality of #1192

miri64 · 2014-06-24T12:04:32Z

And addressed comments in #837

OlegHahm · 2014-06-24T14:50:16Z

tests/test_cpu_id/main.c

@@ -0,0 +1,41 @@
+/*
+ * Copyright (C) 2013 Freie Universität Berlin


OlegHahm · 2014-06-30T09:04:37Z

re-ACK

LudwigKnuepfer · 2014-06-30T09:20:53Z

Maybe you can refresh my memory - why is GET_CPU_ID realized as a macro?

If it really should be implemented as a macro I'd opt for the documentation to indicate that it can't be trusted.
And wouldn't it be good to at least warn if it is used?

haukepetersen · 2014-07-03T21:06:44Z

Just was thinking the same thing as @LudwigOrtmann: what reason did we have to implement it as a macro, and why should it be part of the kernel?

Now that I look at it: wouldn't it make more sense to specify a driver interface that reads out the cpu_id?

miri64 · 2014-07-07T08:20:38Z

I think my reasoning behind using a macro was that it mostly is just the reading of some registers, but true: this can also be done in a function. For the driver: You might be right. I'll look into it.

LudwigKnuepfer · 2014-07-26T15:50:59Z

You a-lookin' yet?

miri64 · 2014-07-26T21:22:48Z

Since there are more pressing issues right now, I postponed that. ;-)

But the general idea is: yes it will be some sort of cpu or random-source "driver".

LudwigKnuepfer · 2014-07-27T05:14:45Z

I just thought the only thing left to do was creating this function..

LudwigKnuepfer · 2014-07-27T06:04:01Z

... and this is a fix me first ...

miri64 · 2014-07-28T08:37:56Z

I discussed this with @haukepetersen and we both agreed that it would make much more sence to put this function into a driver header instead of kernel.h.

Apart from that, the more pressing issues I'm working are also fix me firsts and are much more needed functionality (working ND, working Border Router) ;-)

miri64 · 2014-07-28T08:38:44Z

You can close this PR, if it irritates you...

miri64 · 2014-07-31T18:12:28Z

getter is now a function in its own driver module

LudwigKnuepfer · 2014-07-31T21:01:37Z

You made Travis cry.

haukepetersen · 2014-07-31T21:14:45Z

cpu/native/include/cpu-conf.h

@@ -57,4 +57,6 @@
 /* for nativenet */
 #define NATIVE_ETH_PROTO 0x1234

+#define CPUID_ID_LEN                    (4)


I would suggest a comment on this define.

LudwigKnuepfer · 2014-07-31T21:17:47Z

Æck!

haukepetersen · 2014-07-31T21:20:05Z

ACK

drivers/cpu: add function to get CPU id/serial number

miri64 mentioned this pull request Mar 4, 2014

stm32f103rey6: Define CPU id related macros for stm32f103rey6 OlegHahm/thirdparty_cpu#19

Merged

miri64 assigned LudwigKnuepfer Mar 4, 2014

miri64 added native labels Mar 4, 2014

miri64 mentioned this pull request Mar 4, 2014

auto_init: Initialize link-layer addresses from CPU ID #837

Merged

LudwigKnuepfer reviewed Mar 4, 2014
View reviewed changes

mehlis reviewed Mar 4, 2014
View reviewed changes

LudwigKnuepfer reviewed Mar 5, 2014
View reviewed changes

Kijewski reviewed Mar 7, 2014
View reviewed changes

OlegHahm reviewed Jun 24, 2014
View reviewed changes

tests/test_cpu_id/main.c

@@ -0,0 +1,41 @@

/*

* Copyright (C) 2013 Freie Universität Berlin

Copy link

Member

OlegHahm Jun 24, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

2014

Implement CPU serial number getter.

3a6adcc

miri64 mentioned this pull request Jul 31, 2014

cpu: Provide CPU ID for all supported CPUs #1511

Closed

21 tasks

Implement CPU id getter for native

239f2e0

haukepetersen reviewed Jul 31, 2014
View reviewed changes

miri64 changed the title ~~core: add function to get CPU id/serial number~~ drivers/cpu: add function to get CPU id/serial number Jul 31, 2014

miri64 added a commit that referenced this pull request Jul 31, 2014

Merge pull request #854 from authmillenon/cpu-id

7516c94

drivers/cpu: add function to get CPU id/serial number

miri64 merged commit 7516c94 into RIOT-OS:master Jul 31, 2014

miri64 mentioned this pull request Jul 31, 2014

native: cpu: Add documentatiton for CPUID_ID_LEN #1520

Merged

miri64 deleted the cpu-id branch August 14, 2014 07:35

		@@ -0,0 +1,41 @@
		/*
		* Copyright (C) 2013 Freie Universität Berlin

drivers/cpu: add function to get CPU id/serial number #854

drivers/cpu: add function to get CPU id/serial number #854

Conversation

miri64 commented Mar 4, 2014

LudwigKnuepfer commented Mar 4, 2014

LudwigKnuepfer commented Mar 4, 2014

LudwigKnuepfer commented Mar 4, 2014

OlegHahm commented Mar 4, 2014

miri64 commented Mar 4, 2014

miri64 commented Mar 4, 2014

OlegHahm commented Mar 4, 2014

Choose a reason for hiding this comment

miri64 commented Mar 4, 2014

kaspar030 commented Mar 4, 2014

Choose a reason for hiding this comment

Choose a reason for hiding this comment

miri64 commented Mar 5, 2014

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

miri64 commented Mar 6, 2014

LudwigKnuepfer commented Mar 7, 2014

miri64 commented Mar 7, 2014

Choose a reason for hiding this comment

miri64 commented Mar 24, 2014

thomaseichinger commented May 5, 2014

miri64 commented May 6, 2014

Kijewski commented May 6, 2014

LudwigKnuepfer commented May 13, 2014

OlegHahm commented Jun 22, 2014

miri64 commented Jun 23, 2014

miri64 commented Jun 24, 2014

miri64 commented Jun 24, 2014

Choose a reason for hiding this comment

OlegHahm commented Jun 30, 2014

LudwigKnuepfer commented Jun 30, 2014

haukepetersen commented Jul 3, 2014

miri64 commented Jul 7, 2014

LudwigKnuepfer commented Jul 26, 2014

miri64 commented Jul 26, 2014

LudwigKnuepfer commented Jul 27, 2014

LudwigKnuepfer commented Jul 27, 2014

miri64 commented Jul 28, 2014

miri64 commented Jul 28, 2014

miri64 commented Jul 31, 2014

LudwigKnuepfer commented Jul 31, 2014

Choose a reason for hiding this comment

LudwigKnuepfer commented Jul 31, 2014

haukepetersen commented Jul 31, 2014