Messages in this thread | | | From | "Anvin, H Peter" <> | Date | Fri, 21 Jan 2011 08:06:54 -0800 | Subject | RE: Linux 2.6.38-rc1 doesn't boot |
| |
That is surprisingly painful, which is why the high frequency of binutils bugs is a major headache to us in the kernel world. Either way the current code is just hacky.
-- Sent with Good (www.good.com)
-----Original Message----- From: Lu, Hongjiu Sent: Friday, January 21, 2011 07:37 AM Pacific Standard Time To: Anvin, H Peter; Li, Shaohua Cc: Ingo Molnar; Markus Trippelsdorf; Linus Torvalds; Linux Kernel Mailing List; Sam Ravnborg Subject: RE: Linux 2.6.38-rc1 doesn't boot
Or we can just ban those broken linker versions.
H.J.
> -----Original Message----- > From: Anvin, H Peter > Sent: Friday, January 21, 2011 7:29 AM > To: Li, Shaohua > Cc: Ingo Molnar; Markus Trippelsdorf; Lu, Hongjiu; Linus Torvalds; > Linux Kernel Mailing List; Sam Ravnborg > Subject: Re: Linux 2.6.38-rc1 doesn't boot > > On 01/20/2011 11:18 PM, Li, Shaohua wrote: > > *(.jiffies) > > } > > jiffies = VVIRT(.jiffies); > > + jiffies_64 = jiffies; > > > > OK, this is just messed up. If we need jiffies in a special section > then we should declare it as such (see __jiffy_data), but it's not > clear > to me why we would. This seems like hack upon hack. > > However, I suspect we may want to put jiffies into a .S file, and that > .S file needs to take into account architectures which need underscore > prefixes (which really sucks to have in ELF and is a major fail on > those > ABI designers...) > > -hpa
| |