Skip to contentSkip to conversation

please pull powerpc.git 'merge' branch

PATCH New Paul Mackerras 4 participants 15 messages Apr 29, 2006 – Apr 30, 2006 PowerPC

lore
Share & tools
lore
#0 lore raw

Linus,

Please do a pull from the "merge" branch of

git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc.git

to get the following powerpc updates. They are mostly bug-fixes; there is also a defconfig update for Cell, plus a couple of commits that will enable the kernel to run on forthcoming IBM machines. These don't affect anything on the current machines, since the firmware properties and methods they use aren't present.

Thanks,
Paul.

Alan Modra:
powerpc64: Fix loading of modules without a .toc section

Andreas Schwab:
sound/ppc: snd_pmac_toonie_init should be __init
powerpc: Wire up *at syscalls

Anton Blanchard:
powerpc: Add cputable entry for POWER6

Arnd Bergmann:
spufs: Disable local interrupts for SPE hash_page calls.
powerpc: update cell_defconfig

David Gibson:
powerpc: Fix pagetable bloat for hugepages

David Woodhouse:
powerpc: Use check_legacy_ioport() on ppc32 too.

Eugene Surovegin:

      ppc32: add 440GX erratum 440_43 workaround

Paul Mackerras:
powerpc/pseries: Tell firmware our capabilities on new machines

Vitaly Bordug:
ppc32: odd fixes and improvements in ppc_sys ppc32 CPM_UART: Convert to use platform devices
ppc32: Update board-specific code of the CPM UART users ppc32 CPM_UART: Fixed odd address translations

 arch/powerpc/configs/cell_defconfig         |   40 ++--
 arch/powerpc/kernel/cputable.c              |   16 +
 arch/powerpc/kernel/module_64.c             |   16 +
 arch/powerpc/kernel/prom_init.c             |  112 ++++++++++
 arch/powerpc/kernel/setup-common.c          |    8 +
 arch/powerpc/kernel/setup_64.c              |    8 -
 arch/powerpc/kernel/systbl.S                |   13 +
 arch/powerpc/mm/hugetlbpage.c               |  295 ++++++++++++++++++++++++---
 arch/powerpc/mm/init_64.c                   |    7 +
 arch/powerpc/platforms/cell/spu_base.c      |   10 +
 arch/powerpc/platforms/cell/spu_callbacks.c |   13 +
 arch/ppc/platforms/4xx/ocotea.c             |    2 
 arch/ppc/platforms/mpc8272ads_setup.c       |  114 ++++++++++
 arch/ppc/platforms/mpc866ads_setup.c        |  140 +++++++++++++
 arch/ppc/platforms/mpc885ads_setup.c        |  131 ++++++++++++
 arch/ppc/platforms/pq2ads.c                 |   31 +++
 arch/ppc/syslib/ibm440gx_common.c           |   13 +
 arch/ppc/syslib/ibm440gx_common.h           |    4 
 arch/ppc/syslib/mpc8xx_devices.c            |   25 ++
 arch/ppc/syslib/ppc_sys.c                   |    4 
 arch/ppc/syslib/pq2_sys.c                   |    8 -
 drivers/block/floppy.c                      |    2 
 drivers/input/serio/i8042-io.h              |    4 
 drivers/serial/cpm_uart/cpm_uart.h          |   49 ++++
 drivers/serial/cpm_uart/cpm_uart_core.c     |  280 ++++++++++++++++++++------
 drivers/serial/cpm_uart/cpm_uart_cpm1.c     |   54 -----
 drivers/serial/cpm_uart/cpm_uart_cpm2.c     |   14 -
 fs/stat.c                                   |    2 
 include/asm-powerpc/cputable.h              |   14 +
 include/asm-powerpc/io.h                    |    6 -
 include/asm-powerpc/page_64.h               |    1 
 include/asm-powerpc/pgalloc.h               |    2 
 include/asm-powerpc/unistd.h                |   20 ++
 include/asm-ppc/ppc_sys.h                   |    2 
 include/asm-ppc/reg_booke.h                 |    1 
 sound/ppc/toonie.c                          |    2 
 36 files changed, 1234 insertions(+), 229 deletions(-)
#1 lore raw
Quoted 2 lines · Paul Mackerras

Anton Blanchard:
powerpc: Add cputable entry for POWER6

I never saw this posted to the list.

+#define PPC_FEATURE_ARCH_2_05 0x00001000

Previously we've said implementation instead of specification ("PPC_FEATURE_POWER5_PLUS" etc). That's better than saying base architecture version, since there are parts of the arch that might or might not be implemented (i.e. optional features, etc).

Quoted 2 lines

Paul Mackerras:
powerpc/pseries: Tell firmware our capabilities on new machines

I never saw this one on the list either.

Don't you want to fall back to the ELF method if the prom call fails (ret != 0)? Right you close and return.

-Olof

#2 lore raw
Quoted 4 lines · Olof Johansson

Previously we've said implementation instead of specification ("PPC_FEATURE_POWER5_PLUS" etc). That's better than saying base architecture version, since there are parts of the arch that might or might not be implemented (i.e. optional features, etc).

We now have the AT_PLATFORM string, which we didn't have when we added the POWER5_PLUS etc. features. That specifies which particular implementation we are on quite precisely. We don't want to have a bit for every single implementation or we'll run out of bits.

The ARCH_2_05 bit means all the non-optional bits of the 2.05 architecture. If there are optional features in the architecture, we have separate bits for them. For example, we don't have separate bits for POWER4 and for 970; instead we have a HAS_ALTIVEC bit, and for 970 we set both POWER4 and HAS_ALTIVEC. So the POWER4 bit is really a "2.00 architecture version" bit.

Quoted 2 lines

Don't you want to fall back to the ELF method if the prom call fails (ret != 0)? Right you close and return.

There are two return values here - the return from call-method, and the return from ibm,client-architecture-support. If the latter does not exist, call-method returns non-zero, which gets returned from call_prom_ret, and we close the root node and try the elf-header method. If the ibm,client-architecture-support method exists but returns an error, call_prom_ret returns zero but puts a non-zero value in ret. In that case we don't want to try the elf-header method.

So I claim that the code is correct as it is. :)

Paul.

#3 lore raw
Quoted 10 lines
Quoted 4 lines

Previously we've said implementation instead of specification ("PPC_FEATURE_POWER5_PLUS" etc). That's better than saying base architecture version, since there are parts of the arch that might or might not be implemented (i.e. optional features, etc).

We now have the AT_PLATFORM string, which we didn't have when we added the POWER5_PLUS etc. features. That specifies which particular implementation we are on quite precisely. We don't want to have a bit for every single implementation or we'll run out of bits.

Quoted 6 lines

The ARCH_2_05 bit means all the non-optional bits of the 2.05 architecture. If there are optional features in the architecture, we have separate bits for them. For example, we don't have separate bits for POWER4 and for 970; instead we have a HAS_ALTIVEC bit, and for 970 we set both POWER4 and HAS_ALTIVEC. So the POWER4 bit is really a "2.00 architecture version" bit.

Except that the 970 at least is actually version 2.01. Sounds like names are better than bare numbers -- you showed yourself that numbers like this are confusing, but you also say that a cpu name really means an arch version (which isn't so bad -- there generally is only one "main" CPU per arch version anyway, and people generally know what is compatible to what).

So please rename ARCH_2_05 to the name of the first CPU that implements architecture 2.05?

Segher

#4 lore raw
Quoted 10 lines · Segher Boessenkool
Quoted 10 lines
Quoted 4 lines

Previously we've said implementation instead of specification ("PPC_FEATURE_POWER5_PLUS" etc). That's better than saying base architecture version, since there are parts of the arch that might or might not be implemented (i.e. optional features, etc).

We now have the AT_PLATFORM string, which we didn't have when we added the POWER5_PLUS etc. features. That specifies which particular implementation we are on quite precisely. We don't want to have a bit for every single implementation or we'll run out of bits.

Quoted 8 lines
Quoted 6 lines

The ARCH_2_05 bit means all the non-optional bits of the 2.05 architecture. If there are optional features in the architecture, we have separate bits for them. For example, we don't have separate bits for POWER4 and for 970; instead we have a HAS_ALTIVEC bit, and for 970 we set both POWER4 and HAS_ALTIVEC. So the POWER4 bit is really a "2.00 architecture version" bit.

Except that the 970 at least is actually version 2.01.

Are there any userspace-visible differences between 2.00 and 2.01?

Quoted 6 lines

Sounds
like names are better than bare numbers -- you showed yourself that numbers like this are confusing, but you also say that a cpu name really means an arch version (which isn't so bad -- there generally is only one "main" CPU per arch version anyway, and people generally know what is compatible to what).

It's actually about an even tradeoff between the two. We run the risk of running out of bits eventually no matter what, and as long as it's blindingly clear and well stated that it does not imply any optional features, then it doesn't matter. This puts the burden on userspace developers to know which features are optional, not just what works on their POWER6 lab machine, etc.

Quoted 2 lines

So please rename ARCH_2_05 to the name of the first CPU that implements architecture 2.05?

We already have SMT, ICACHE_SNOOP and ALTIVEC feature bits, so it probably makes sense to stay in the ARCH_2_05 direction. Unfortunately it might be tricky to rename the old defines now due to userland exposure.

I guess Cell breaks all this logic though. Which PPC base version does it implement?

-Olof

#5 lore raw
Quoted 6 lines · Olof Johansson

We already have SMT, ICACHE_SNOOP and ALTIVEC feature bits, so it probably makes sense to stay in the ARCH_2_05 direction. Unfortunately it might be tricky to rename the old defines now due to userland exposure.

I guess Cell breaks all this logic though. Which PPC base version does it implement?

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates some of the extensions that are optional in there.
The relevant sections in the specs are:

| 4.1.1 Optional Features in PowerPC Architecture, Book I (Required for
|       CBEA) 
| The following facilities and instructions are considered optional in
| the PowerPC Architecture, but are required for the PPE by the CBEA
| user mode environment. 
|  • Floating reciprocal estimate single A-form (fres)
|  • Floating reciprocal square-root estimate A-form (frsqte)
|  • Vector/SIMD multimedia extension
| Note: The optional PowerPC floating-point instructions that are
| mandatory in the CBEA are needed for the application space targeted
| by the CBEA. 
|
| 4.1.3 Optional Features in PowerPC Architecture, Book II (Required
|       for CBEA) 
| The following facilities and instructions are considered optional
| in the PowerPC Architecture, but are required in the CBEA.
|  • Data cache block touch X-form (dcbt)
|    This is an optional version of dcbt that permits a program to
|    provide a hint that a sequence of data cache blocks is likely
|    to be needed soon. 
|  Book II describes these facilities and instructions.

I believe the actual implementation is missing one of the mandatory features of 2.03 that was not in 2.02 though.

	Arnd <><
#6 lore raw
Quoted 2 lines · Arnd Bergmann

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates some of the extensions that are optional in there.

Oh, and it also has some extensions that are supposed to be part of later powerpc architecture releases (CBEA, appendix E):

• Mediated interrupt
• Multiple concurrent large pages
• Add TL (TLB load control) bit to logical-partitioning control register (LPCR) for software versus hardware load of TLB
• Change tlbie<l> of large page to always invalidate ERAT
• Hypervisor SPRs no longer readable when HV equals 0

Do we know which ones of these made it into 2.05?

	Arnd <><
#7 lore raw
Quoted 8 lines · Arnd Bergmann

• Mediated interrupt
• Multiple concurrent large pages
• Add TL (TLB load control) bit to logical-partitioning control register (LPCR) for software versus hardware load of TLB
• Change tlbie<l> of large page to always invalidate ERAT
• Hypervisor SPRs no longer readable when HV equals 0

Do we know which ones of these made it into 2.05?

I'm not sure that can be discussed on a public mailing list, given that 2.02 is the last non-secret architecture version at this time. :-)

-Olof

#8 lore raw
In reply to Arnd Bergmann
Quoted 8 lines
Quoted 6 lines

We already have SMT, ICACHE_SNOOP and ALTIVEC feature bits, so it probably makes sense to stay in the ARCH_2_05 direction. Unfortunately it might be tricky to rename the old defines now due to userland exposure.

I guess Cell breaks all this logic though. Which PPC base version does it implement?

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates

The docs (for example, "CBE_Architecture_v10.pdf") say 2.02. One more reason to not use those arch version numbers.

Segher

#9 lore raw
Quoted 7 lines · Segher Boessenkool
Quoted 3 lines

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates

The docs (for example, "CBE_Architecture_v10.pdf") say 2.02.  One more reason
to not use those arch version numbers.

Sorry, my fault. I misremembered it from that spec.

	Arnd <><
#10 lore raw
In reply to Arnd Bergmann
Quoted 3 lines · Arnd Bergmann

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates some of the extensions that are optional in there.
The relevant sections in the specs are:

Ok. So Cell should really set PPC_FEATURE_ARCH_2_02 then, and possible PPC_FEATURE_CELL to signify the cell extensions?

Right now ARCH_2_02 is not set as far as I can see.

-Olof

#11 lore raw
Quoted 10 lines · Olof Johansson
Quoted 3 lines · Arnd Bergmann

The Cell Broadband Engine Architecture in theory is 2.03 and also mandates some of the extensions that are optional in there.
The relevant sections in the specs are:

Ok. So Cell should really set PPC_FEATURE_ARCH_2_02 then, and possible PPC_FEATURE_CELL to signify the cell extensions?

Right now ARCH_2_02 is not set as far as I can see.

s/2_02/2_03/g

-Olof

#12 lore raw
In reply to Paul Mackerras
Quoted 12 lines · Paul Mackerras
Quoted 4 lines · Olof Johansson

Previously we've said implementation instead of specification ("PPC_FEATURE_POWER5_PLUS" etc). That's better than saying base architecture version, since there are parts of the arch that might or might not be implemented (i.e. optional features, etc).

We now have the AT_PLATFORM string, which we didn't have when we added the POWER5_PLUS etc. features. That specifies which particular implementation we are on quite precisely. We don't want to have a bit for every single implementation or we'll run out of bits.

Quoted 6 lines

The ARCH_2_05 bit means all the non-optional bits of the 2.05 architecture. If there are optional features in the architecture, we have separate bits for them. For example, we don't have separate bits for POWER4 and for 970; instead we have a HAS_ALTIVEC bit, and for 970 we set both POWER4 and HAS_ALTIVEC. So the POWER4 bit is really a "2.00 architecture version" bit.

Ok, I guess that's more or less how it's been used, since there's only been one implementation per architecture version in the past (970 series being the main exception). The only one left out from the table seems to be 2.01 (POWER4+), not sure if it had any userspace-visible changes in the first place.

A rename of existing names might make sense for consistency/clarity. But that's outside of __KERNEL__ so it would break userspace.

Quoted 6 lines
Quoted 2 lines

Don't you want to fall back to the ELF method if the prom call fails (ret != 0)? Right you close and return.

If the ibm,client-architecture-support method exists but returns an error, call_prom_ret returns zero but puts a non-zero value in ret.

Right, that's quite obvious from the implementation.

Quoted 1 line

In that case we don't want to try the elf-header method.

That was my question, if it made sense to fall back to it. And that's a "no". Thanks. :-)

-Olof

#13 lore raw

PATCH 1/1 powerpc: rename PPC_FEATURE_POWER.* to PPC_FEATURE_ARCH.*

Quoted 2 lines · Olof Johansson

A rename of existing names might make sense for consistency/clarity. But that's outside of __KERNEL__ so it would break userspace.

Less talk, more patches. 2.6.18 material.

Btw, looks like glibc duplicates the definitions instead of includes them, but let's be nice and keep them anyway. Besides, they got the ISA version to processor mappings wrong in their comments.

---

Rename PPC_FEATURE_POWER.* to be signify the base architecture versions instead. Keep the old names around since they're defined outside of __KERNEL__.

If POWER5 is 2.02 (last public version), and POWER6 is 2.05, then POWER5+ should be 2.03 or 2.04. Not sure what fills in the hole, but I took a guess that 5+ is 2.03.

arch/powerpc/kernel/cputable.c +3 -3
index 3f7182d..7017d1c 100644
@@ -52,10 +52,10 @@ extern void __setup_cpu_ppc970(unsigned 
 #define COMMON_USER		(PPC_FEATURE_32 | PPC_FEATURE_HAS_FPU | \
 				 PPC_FEATURE_HAS_MMU)
 #define COMMON_USER_PPC64	(COMMON_USER | PPC_FEATURE_64)
-#define COMMON_USER_POWER4	(COMMON_USER_PPC64 | PPC_FEATURE_POWER4)
-#define COMMON_USER_POWER5	(COMMON_USER_PPC64 | PPC_FEATURE_POWER5 |\
+#define COMMON_USER_POWER4	(COMMON_USER_PPC64 | PPC_FEATURE_ARCH_2_00)
+#define COMMON_USER_POWER5	(COMMON_USER_PPC64 | PPC_FEATURE_ARCH_2_02 |\
 				 PPC_FEATURE_SMT | PPC_FEATURE_ICACHE_SNOOP)
-#define COMMON_USER_POWER5_PLUS	(COMMON_USER_PPC64 | PPC_FEATURE_POWER5_PLUS|\
+#define COMMON_USER_POWER5_PLUS	(COMMON_USER_PPC64 | PPC_FEATURE_ARCH_2_03|\
 				 PPC_FEATURE_SMT | PPC_FEATURE_ICACHE_SNOOP)
 #define COMMON_USER_POWER6	(COMMON_USER_PPC64 | PPC_FEATURE_ARCH_2_05 |\
 				 PPC_FEATURE_SMT | PPC_FEATURE_ICACHE_SNOOP)
include/asm-powerpc/cputable.h +12 -3
index 9fcf016..deae3af 100644
@@ -15,15 +15,24 @@
 #define PPC_FEATURE_HAS_EFP_SINGLE	0x00400000
 #define PPC_FEATURE_HAS_EFP_DOUBLE	0x00200000
 #define PPC_FEATURE_NO_TB		0x00100000
-#define PPC_FEATURE_POWER4		0x00080000
-#define PPC_FEATURE_POWER5		0x00040000
-#define PPC_FEATURE_POWER5_PLUS		0x00020000
+#define PPC_FEATURE_ARCH_2_00		0x00080000
+#define PPC_FEATURE_ARCH_2_02		0x00040000
+#define PPC_FEATURE_ARCH_2_03		0x00020000
 #define PPC_FEATURE_CELL		0x00010000
 #define PPC_FEATURE_BOOKE		0x00008000
 #define PPC_FEATURE_SMT			0x00004000
 #define PPC_FEATURE_ICACHE_SNOOP	0x00002000
 #define PPC_FEATURE_ARCH_2_05		0x00001000
 
+/*
+ * These are needed for userspace compat, since processor
+ * models used to be used before instead of arch versions
+ */
+
+#define PPC_FEATURE_POWER4	PPC_FEATURE_ARCH_2_00
+#define PPC_FEATURE_POWER5	PPC_FEATURE_ARCH_2_02
+#define PPC_FEATURE_POWER5_PLUS	PPC_FEATURE_ARCH_2_03
+
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
 
#14 lore raw

PATCH 1/1 powerpc: Fix comment in ibm_architecture_vec

In reply to Paul Mackerras
Quoted 2 lines · Paul Mackerras

Paul Mackerras:
powerpc/pseries: Tell firmware our capabilities on new machines

Below is definitely not 2.6.17 material, so for the 2.6.18 queue:

---

Either constant or comment is wrong, I'm guessing it's comment since the ELF header version uses 64MB.

arch/powerpc/kernel/prom_init.c +1 -1
index 078fb55..ef99f26 100644
@@ -705,7 +705,7 @@ static unsigned char ibm_architecture_ve
 	W(0xffffffff),			/* virt_base */
 	W(0xffffffff),			/* virt_size */
 	W(0xffffffff),			/* load_base */
-	W(64),				/* 128MB min RMA */
+	W(64),				/* 64MB min RMA */
 	W(0xffffffff),			/* full client load */
 	0,				/* min RMA percentage of total RAM */
 	48,				/* max log_2(hash table size) */