netdev - Re: [patch iproute2-next v3 3/6] devlink: extend pr_out_nested

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <871qdc5mcj.fsf@nvidia.com>
Date: Mon, 30 Oct 2023 12:03:57 +0100
From: Petr Machata <petrm@...dia.com>
To: David Ahern <dsahern@...il.com>
CC: Petr Machata <me@...chata.org>, Jiri Pirko <jiri@...nulli.us>,
	<netdev@...r.kernel.org>, <stephen@...workplumber.org>,
	<daniel.machon@...rochip.com>
Subject: Re: [patch iproute2-next v3 3/6] devlink: extend
 pr_out_nested_handle() to print object


David Ahern <dsahern@...il.com> writes:

> On 10/27/23 7:12 AM, Petr Machata wrote:
>> I was wondering whether somehing like this might make sense in the
>> iproute2 library:
>> 
>> 	#define alloca_sprintf(FMT, ...) ({					\
>> 		int xasprintf_n = snprintf(NULL, 0, (FMT), __VA_ARGS__);	\
>> 		char *xasprintf_buf = alloca(xasprintf_n);			\
>> 		sprintf(xasprintf_buf, (FMT), __VA_ARGS__);			\
>> 		xasprintf_buf;							\
>> 	})
>> 
>> 	void foo() {
>> 		const char *buf = alloca_sprintf("%x %y %z", etc.);
>> 		printf(... buf ...);
>> 	}
>> 
>> I'm not really happy with it -- because of alloca vs. array, and because
>> of the double evaluation. But all those SPRINT_BUF's peppered everywhere
>> make me uneasy every time I read or write them.
>
> agreed.
>
>> 
>> Or maybe roll something custom asprintf-like that can reuse and/or
>> realloc a passed-in buffer?
>> 
>> The sprintf story is pretty bad in iproute2 right now, IMHO.
>
> It is a bit of a mess. If you have a few cycles, want to send an RFC?
> Just pick 1 or 2 to convert to show intent with a new design.

I picked at it a bit over the weekend, but came up with nothing that I
find comfortable proposing. The static buffer approach has some major
advantages: nothing ever fails and nothing ever needs cleanups. This
keeps the client code tidy and compact. Anything dynamic adds points of
failure and cleanups, which in C means more client-side boilerplate.
Anyway, I'll pick at it some more and see I find anything.