Re: [PATCH 00/13] Remove metag architecture
On Thu, Feb 22, 2018 at 12:38 AM, James Hoganwrote: > These patches remove the metag architecture and tightly dependent > drivers from the kernel. With the 4.16 kernel the ancient gcc 4.2.4 > based metag toolchain we have been using is hitting compiler bugs, so > now seems a good time to drop it altogether. > > Quoting from patch 1: > > The earliest Meta architecture port of Linux I have a record of was an > import of a Meta port of Linux v2.4.1 in February 2004, which was worked > on significantly over the next few years by Graham Whaley, Will Newton, > Matt Fleming, myself and others. > > Eventually the port was merged into mainline in v3.9 in March 2013, not > long after Imagination Technologies bought MIPS Technologies and shifted > its CPU focus over to the MIPS architecture. > > As a result, though the port was maintained for a while, kept on life > support for a while longer, and useful for testing a few specific > drivers for which I don't have ready access to the equivalent MIPS > hardware, it is now essentially dead with no users. > > It is also stuck using an out-of-tree toolchain based on GCC 4.2.4 which > is no longer maintained, now struggles to build modern kernels due to > toolchain bugs, and doesn't itself build with a modern GCC. The latest > buildroot port is still using an old uClibc snapshot which is no longer > served, and the latest uClibc doesn't build with GCC 4.2.4. > > So lets call it a day and drop the Meta architecture port from the > kernel. RIP Meta. I've pulled it into my asm-generic tree now, which is also part of linux-next, and followed up with patches removing frv, m32r, score, unicore32 and blackfin. I have not removed the device drivers yet, but I'm working on that. Arnd
Re: [PATCH 00/13] Remove metag architecture
On Fri, Feb 23, 2018 at 12:02 PM, James Hoganwrote: > On Fri, Feb 23, 2018 at 11:26:58AM +0100, Arnd Bergmann wrote: >> On Thu, Feb 22, 2018 at 12:38 AM, James Hogan wrote: >> > So lets call it a day and drop the Meta architecture port from the >> > kernel. RIP Meta. >> >> Since I brought up the architecture removal independently, I could >> pick this up into a git tree that also has the removal of some of the >> other architectures. >> >> I see your tree is part of linux-next, so you could also just put it >> in there and send a pull request at the merge window if you prefer. >> >> The only real reason I see for a shared git tree would be to avoid >> conflicts when we touch the same Kconfig files or #ifdefs in driver, >> but Meta only appears in >> >> config FRAME_POINTER >> bool "Compile the kernel with frame pointers" >> depends on DEBUG_KERNEL && \ >> (CRIS || M68K || FRV || UML || \ >> SUPERH || BLACKFIN || MN10300 || METAG) || \ >> ARCH_WANT_FRAME_POINTERS >> >> and >> >> include/trace/events/mmflags.h:#elif defined(CONFIG_PARISC) || >> defined(CONFIG_METAG) || defined(CONFIG_IA64) >> >> so there is little risk. > > I'm happy to put v2 in linux-next now (only patch 4 has changed, I just > sent an updated version), and send you a pull request early next week so > you can take it from there. The patches can't be directly applied with > git-am anyway thanks to the -D option to make them more concise. > > Sound okay? Yes, sounds good, thanks! Arnd
Re: [PATCH 00/13] Remove metag architecture
On Fri, Feb 23, 2018 at 11:26:58AM +0100, Arnd Bergmann wrote: > On Thu, Feb 22, 2018 at 12:38 AM, James Hoganwrote: > > So lets call it a day and drop the Meta architecture port from the > > kernel. RIP Meta. > > Since I brought up the architecture removal independently, I could > pick this up into a git tree that also has the removal of some of the > other architectures. > > I see your tree is part of linux-next, so you could also just put it > in there and send a pull request at the merge window if you prefer. > > The only real reason I see for a shared git tree would be to avoid > conflicts when we touch the same Kconfig files or #ifdefs in driver, > but Meta only appears in > > config FRAME_POINTER > bool "Compile the kernel with frame pointers" > depends on DEBUG_KERNEL && \ > (CRIS || M68K || FRV || UML || \ > SUPERH || BLACKFIN || MN10300 || METAG) || \ > ARCH_WANT_FRAME_POINTERS > > and > > include/trace/events/mmflags.h:#elif defined(CONFIG_PARISC) || > defined(CONFIG_METAG) || defined(CONFIG_IA64) > > so there is little risk. I'm happy to put v2 in linux-next now (only patch 4 has changed, I just sent an updated version), and send you a pull request early next week so you can take it from there. The patches can't be directly applied with git-am anyway thanks to the -D option to make them more concise. Sound okay? Thanks James signature.asc Description: Digital signature
Re: [PATCH 00/13] Remove metag architecture
On Thu, Feb 22, 2018 at 12:38 AM, James Hoganwrote: > These patches remove the metag architecture and tightly dependent > drivers from the kernel. With the 4.16 kernel the ancient gcc 4.2.4 > based metag toolchain we have been using is hitting compiler bugs, so > now seems a good time to drop it altogether. > > Quoting from patch 1: > > The earliest Meta architecture port of Linux I have a record of was an > import of a Meta port of Linux v2.4.1 in February 2004, which was worked > on significantly over the next few years by Graham Whaley, Will Newton, > Matt Fleming, myself and others. > > Eventually the port was merged into mainline in v3.9 in March 2013, not > long after Imagination Technologies bought MIPS Technologies and shifted > its CPU focus over to the MIPS architecture. > > As a result, though the port was maintained for a while, kept on life > support for a while longer, and useful for testing a few specific > drivers for which I don't have ready access to the equivalent MIPS > hardware, it is now essentially dead with no users. > > It is also stuck using an out-of-tree toolchain based on GCC 4.2.4 which > is no longer maintained, now struggles to build modern kernels due to > toolchain bugs, and doesn't itself build with a modern GCC. The latest > buildroot port is still using an old uClibc snapshot which is no longer > served, and the latest uClibc doesn't build with GCC 4.2.4. > > So lets call it a day and drop the Meta architecture port from the > kernel. RIP Meta. Since I brought up the architecture removal independently, I could pick this up into a git tree that also has the removal of some of the other architectures. I see your tree is part of linux-next, so you could also just put it in there and send a pull request at the merge window if you prefer. The only real reason I see for a shared git tree would be to avoid conflicts when we touch the same Kconfig files or #ifdefs in driver, but Meta only appears in config FRAME_POINTER bool "Compile the kernel with frame pointers" depends on DEBUG_KERNEL && \ (CRIS || M68K || FRV || UML || \ SUPERH || BLACKFIN || MN10300 || METAG) || \ ARCH_WANT_FRAME_POINTERS and include/trace/events/mmflags.h:#elif defined(CONFIG_PARISC) || defined(CONFIG_METAG) || defined(CONFIG_IA64) so there is little risk. Arnd
Re: [PATCH 00/13] Remove metag architecture
On 22 February 2018 at 09:48, James Hoganwrote: > On Thu, Feb 22, 2018 at 10:26:54AM +0100, Peter Zijlstra wrote: >> On Wed, Feb 21, 2018 at 11:38:12PM +, James Hogan wrote: >> > So lets call it a day and drop the Meta architecture port from the >> > kernel. RIP Meta. >> >> So long, and thanks for all the fish! >> >> Nice cleanup though, most welcome :-) > > I thought you might like it ;-) > >> Acked-by: Peter Zijlstra (Intel) > > Thanks > James RIP indeed. As I sit here listening to my META powered radio for the series in general: Acked-by: Graham Whaley
Re: [PATCH 00/13] Remove metag architecture
On Thu, Feb 22, 2018 at 10:26:54AM +0100, Peter Zijlstra wrote: > On Wed, Feb 21, 2018 at 11:38:12PM +, James Hogan wrote: > > So lets call it a day and drop the Meta architecture port from the > > kernel. RIP Meta. > > So long, and thanks for all the fish! > > Nice cleanup though, most welcome :-) I thought you might like it ;-) > Acked-by: Peter Zijlstra (Intel)Thanks James signature.asc Description: Digital signature
Re: [PATCH 00/13] Remove metag architecture
On Wed, Feb 21, 2018 at 11:38:12PM +, James Hogan wrote: > So lets call it a day and drop the Meta architecture port from the > kernel. RIP Meta. So long, and thanks for all the fish! Nice cleanup though, most welcome :-) Acked-by: Peter Zijlstra (Intel)