Params: Add parameter types for IP addresses in various forms.

Information
Submitter:	Gabe Black
Repository:	gem5
Branch:
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Params: Add parameter types for IP addresses in various forms.

New parameter forms are:
IP address in the format "a.b.c.d" where a-d are from decimal 0 to 255.
IP address with netmask which is an IP followed by "/n" where n is a netmask
length in bits from decimal 0 to 32 or by "/e.f.g.h" where e-h are from
decimal 0 to 255 and which is all 1 bits followed by all 0 bits when
represented in binary. These can also be specified as an integral IP and
netmask passed in separately.
IP address with port which is an IP followed by ":p" where p is a port index
from decimal 0 to 65535. These can also be specified as an integral IP and
port value passed in separately.

Testing Done

src/python/m5/params.py (Diff revision 1)

All this kwargs stuff seems like overkill, and I'm not sure all the corner cases are handled sensibly (passing in args[0] as elseVal for both ip and netmask seems particularly suspect), but now that it's written I guess there's no point in deleting it.  A comment that lists all the options for specifying a value would be helpful though.

Gabe Black Nov. 17, 2010, 11:37 a.m. (Nov. 17, 2010, 11:37 a.m.)

This is based off of the implementation of the Range param type, and lets you set the ip and port or ip and netmask explicitly. That's very handy when generating those values programmatically which is exactly how I've been using them. You're right that not all cases are covered. If kargs has neither ip nor netmask and one value args, that will end up in both ip and netmask. I'll look it over and see if some of the gaps can be covered easily. I was mimicking the Range type like I mentioned, but maybe that's not the best approach.

src/python/m5/util/convert.py (Diff revision 1)

Shouldn't this be "< 256"?  (and "< 32" below too?)  Didn't you just point out this same bug in some existing code here?

Gabe Black Nov. 17, 2010, 11:24 a.m. (Nov. 17, 2010, 11:24 a.m.)

< 256 yes, < 32 no. The max byte value is 255, but the netmask can have 0 to 32 bits set I think, where 0 would be all possible IPs and 32 would be one specific IP. Let me know if that's not true.

Gabe Black Nov. 17, 2010, 11:43 a.m. (Nov. 17, 2010, 11:43 a.m.)
```
And, yes it's the same thing I pointed out earlier. :-)
```

src/base/inet.cc (Diff revision 1)

This is somewhat backwards as cprintf can take an ostream.  I'd make operator<< do the right thing with ccprintf and then have string() use it.  Just like EthAddr.

Gabe Black Nov. 17, 2010, 11:51 a.m. (Nov. 17, 2010, 11:51 a.m.)

Yeah, doing things this other way around just seemed to go together easier, but I think it's actually a little less efficient too. I'll look at turning it back around.

src/python/m5/params.py (Diff revision 1)

There are already checks in convert for this.  Seems odd to do it here too.

Gabe Black Nov. 17, 2010, 11:50 a.m. (Nov. 17, 2010, 11:50 a.m.)

This is for when the values are set explicitly and don't go through convert. It needs to be in convert as well since that can be called in other contexts. I could move the verification functions into convert.py and call them from both places, but they're so short and simple it doesn't seem like it'd be worth it.

src/python/m5/params.py (Diff revision 1)

There are already checks in convert for this.  Seems odd to have them in both places.

Gabe Black Nov. 17, 2010, 11:50 a.m. (Nov. 17, 2010, 11:50 a.m.)
```
Ditto.
```

Diff:

Revision 2 (+338)

Show changes

	src/base/inet.hh
	src/base/inet.cc
	src/python/m5/params.py
	src/python/m5/util/convert.py
	src/python/swig/inet.i

Diff:

Revision 3 (+338)

Show changes

	src/base/inet.hh
	src/base/inet.cc
	src/python/m5/params.py
	src/python/m5/util/convert.py
	src/python/swig/inet.i

src/python/m5/util/convert.py (Diff revision 3)

I net mask could also be specified like 10.3.4.5/255.255.0.0. Perhaps handling that case would be good too. Minimally, I think the /16 should be converted to the long form netmask and that should be passed around.

Gabe Black Nov. 20, 2010, 2:26 p.m. (Nov. 20, 2010, 2:26 p.m.)

It would be a good idea to recognize the long form, but I think the short form should be passed around. It's a lot easier to reliably go from short to long than the other way.

src/python/m5/util/convert.py (Diff revision 3)

Why hex for max port here and decimal for max byte above?

Gabe Black Nov. 20, 2010, 2:26 p.m. (Nov. 20, 2010, 2:26 p.m.)

It was arbitrary. It looks like everything else uses hex values so I'll standardize on that.

Diff:

Revision 4 (+352)

Show changes

	src/base/inet.hh
	src/base/inet.cc
	src/python/m5/params.py
	src/python/m5/util/convert.py
	src/python/swig/inet.i

Description:

		Params: Add parameter types for IP addresses in various forms.

		New parameter forms are:
		IP address in the format "a.b.c.d" where a-d are from decimal 0 to 255.
		IP address with netmask which is an IP followed by "/n" where n is a netmask
~		length in bits from decimal 0 to 32. These can also be specified as an
~		integral IP and netmask passed in separately.
	~	length in bits from decimal 0 to 32 or by "/e.f.g.h" where e-h are from
	~	decimal 0 to 255 and which is all 1 bits followed by all 0 bits when
	+	represented in binary. These can also be specified as an integral IP and
	+	netmask passed in separately.
		IP address with port which is an IP followed by ":p" where p is a port index
		from decimal 0 to 65535. These can also be specified as an integral IP and
		port value passed in separately.

Diff:

Revision 5 (+352)

Show changes

	src/base/inet.hh
	src/base/inet.cc
	src/python/m5/params.py
	src/python/m5/util/convert.py
	src/python/swig/inet.i

src/base/inet.cc (Diff revision 5)

I don't see how this works. after the first nibble, wont everything be 0?

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Diff:

Diff:

Diff:

Description:

Diff: