linux-kernel - Re: [PATCH 0/4] Speed up the symbols' resolution process V4

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

Message-ID: <BANLkTikFSN-DkkPNPtv1VigCg08YoB_q2g@mail.gmail.com>
Date:	Wed, 11 May 2011 11:52:56 -0400
From:	Mike Frysinger <vapier.adi@...il.com>
To:	Alessio Igor Bogani <abogani@...nel.org>
Cc:	Rusty Russell <rusty@...tcorp.com.au>,
	Tim Abbott <tabbott@...lice.com>,
	Anders Kaseorg <andersk@...lice.com>,
	Jason Wessel <jason.wessel@...driver.com>,
	Tim Bird <tim.bird@...sony.com>,
	LKML <linux-kernel@...r.kernel.org>,
	Linux Embedded <linux-embedded@...r.kernel.org>
Subject: Re: [PATCH 0/4] Speed up the symbols' resolution process V4

On Wed, May 11, 2011 at 11:25, Alessio Igor Bogani wrote:
> 2011/5/11 Mike Frysinger:
>>> Sorry I don't think that is a good choice from a long term point of
>>> view. What do you think to add MODULE_SYMBOL_PREFIX to section names
>>> instead? In this way symbol and section names should always be
>>> different also on symbol prefixed archs (which are blackfin and
>>> h8300).
>>
>> that doesnt work.  it simply delays the problem to another set of
>> underscores.  so with that change, local_bh_enable/_local_bh_enable
>> work, but now send_remote_softirq/__send_remote_softirq fail:
>
> In my opinion it should work. if I use SYMBOL_PREFIX + two underscore
> for section name I should always obtain different names.
> So if SYMBOL_PREFIX is "_" section name will be "___", if
> SYMBOL_PREFIX is "__" section name will be "____" and so on.

i dont follow your logic.  SYMBOL_PREFIX is simply an underscore for
Blackfin, so you are not guaranteed unique names when you're only
prefixing more underscores.  the issue isnt with a symbol colliding
with itself, the issue is with a symbol colliding with another symbol
that has underscores added to its name.

if you export _foo/foo, you'll get an error with the current code:
/* EXPORT_SYMBOL(foo); */
        .section        ___ksymtab__foo,"a",@progbits
___ksymtab_foo:
/* EXPORT_SYMBOL(_foo); */
        .section        ___ksymtab___foo,"a",@progbits
___ksymtab__foo:

you can see how the first section and the last symbol collide

by putting the symbol prefix along the lines of the separator, you
delay it to __foo/foo:
/* EXPORT_SYMBOL(foo); */
        .section        ___ksymtab___foo,"a",@progbits
___ksymtab_foo:
/* EXPORT_SYMBOL(__foo); */
        .section        ___ksymtab_____foo,"a",@progbits
___ksymtab___foo:

the Blackfin toolchain puts the prefix before the ksymtab part while
you're putting it after, so there is always a possibility of collision
unless you pick a split char that doesnt include an underscore.

>>  CC      kernel/softirq.o
>> nano /tmp/cconhYy1.s: Assembler messages:
>> /tmp/cconhYy1.s:3664: Error: symbol `___ksymtab___send_remote_softirq'
>> is already defined
>> make[1]: *** [kernel/softirq.o] Error 1
>
> I'm a bit confused. I can build a kernel here:
> $ make ARCH=blackfin CROSS_COMPILE="bfin-uclinux-" defconfig

it's failing for me with current linux-next and the patch you posted
previously.  also, you dont need to set CROSS_COMPILE as that is the
default value for ARCH=blackfin ...

> Unfortunately I can't make skyeye emulator works to test the obtained
> kernel image.

skyeye is crap.  the latest upstream gdb/sim code can boot a kernel,
but if you aren't up-to-date with that, this binary i've been using
locally should work too:
http://blackfin.uclinux.org/uimages/bfin-elf-run

$ bfin-elf-run --env operating --model bf537 --sirev 2 ./vmlinux
earlyprintk=serial,uart0 console=ttyBF0
Linux version 2.6.39-rc7-ADI-2011R1-pre-00132-g9d91c9a
(vapier@...ier-m) (gcc version 4.3.5 (ADI-trunk/git-ce854ee) ) #16 Wed
May 11 11:55:28 EDT 2011
register early platform devices
bootconsole [early_shadow0] enabled
bootconsole [early_BFuart0] enabled
early printk enabled on early_BFuart0
Limiting kernel memory to 56MB due to anomaly 05000263
Board Memory: 64MB
Kernel Managed Memory: 64MB
Memory map:
  fixedcode = 0x00000400-0x00000490
  text      = 0x00001000-0x000f7f30
  rodata    = 0x000f7f30-0x001456ac
.............
-mike
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/