Skip to contentSkip to conversation

x86: random clean-ups

PATCH In review Dmitri Vorobiev 5 participants 21 messages Apr 20, 2008 – Apr 24, 2008 1 message not in the archive x86

lore
Share & tools
lore
#0 lore raw

cover x86: random clean-ups

Hi Ingo, Peter, Thomas,

The following patches contain a few clean-ups, mostly stemming from analyzis of sparse messages.

The patches are against current x86.git.

Please consider.

Thanks,
Dmitri Vorobiev

#1 lore raw

PATCH 1/4 x86: remove unused function amd_init_cpu()

There are no users for the function amd_init_cpu() defined in arch/x86/kernel/cpu/amd.c. This patch removes this routine.

This patch was build-tested using defconfigs for i386 and x86_64, and a few randconfig instances. Runtime tests were performed by booting 32- and 64-bit x86 boxen up to the shell prompt.

---
 arch/x86/kernel/cpu/amd.c |    6 ------
 1 files changed, 0 insertions(+), 6 deletions(-)
arch/x86/kernel/cpu/amd.c +0 -6
index 0173065..2458668 100644
@@ -343,10 +343,4 @@ static struct cpu_dev amd_cpu_dev __cpuinitdata = {
 	.c_size_cache	= amd_size_cache,
 };
 
-int __init amd_init_cpu(void)
-{
-	cpu_devs[X86_VENDOR_AMD] = &amd_cpu_dev;
-	return 0;
-}
-
 cpu_vendor_dev_register(X86_VENDOR_AMD, &amd_cpu_dev);
Signature · 1 line
1.5.3
#2 lore raw

PATCH 2/4 x86: remove unused code and data in nexgen.c

In reply to Dmitri Vorobiev

Nobody calls the function nexgen_init_cpu(), therefore it can be removed. The variable nexgen_cpu_dev is used only by this function, so this patch removes useless data as well.

This patch was build-tested using defconfigs for i386 and x86_64. It also survived a few randconfig instances for i386. Runtime tests were performed by booting 32-bit and 64-bit x86 boxen up to the shell prompt.

---
 arch/x86/kernel/cpu/nexgen.c |   19 -------------------
 1 files changed, 0 insertions(+), 19 deletions(-)
arch/x86/kernel/cpu/nexgen.c +0 -19
index 5d5e1c1..f5753b4 100644
@@ -38,22 +38,3 @@ static void __cpuinit nexgen_identify(struct cpuinfo_x86 *c)
 	if (deep_magic_nexgen_probe())
 		strcpy(c->x86_vendor_id, "NexGenDriven");
 }
-
-static struct cpu_dev nexgen_cpu_dev __cpuinitdata = {
-	.c_vendor	= "Nexgen",
-	.c_ident	= { "NexGenDriven" },
-	.c_models = {
-			{ .vendor = X86_VENDOR_NEXGEN,
-			  .family = 5,
-			  .model_names = { [1] = "Nx586" }
-			},
-	},
-	.c_init		= init_nexgen,
-	.c_identify	= nexgen_identify,
-};
-
-int __init nexgen_init_cpu(void)
-{
-	cpu_devs[X86_VENDOR_NEXGEN] = &nexgen_cpu_dev;
-	return 0;
-}
Signature · 1 line
1.5.3
#3 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Quoted 8 lines · Dmitri Vorobiev

Nobody calls the function nexgen_init_cpu(), therefore it can be removed. The variable nexgen_cpu_dev is used only by this function, so this patch removes useless data as well.

This patch was build-tested using defconfigs for i386 and x86_64. It also survived a few randconfig instances for i386. Runtime tests were performed by booting 32-bit and 64-bit x86 boxen up to the shell prompt.

NAK. This is broken.

The reason nexgen_cpu_dev is unreferenced is because the file is missing a cpu_vendor_dev_register(), which should be added.

-hpa

#4 lore raw

PATCH v2 2/2 x86: remove unused function, register the cpu vendor in nexgen.c

Nobody calls the function nexgen_init_cpu(), which is defined in arch/x86/kernel/cpu/nexgen.c. Therefore this routine can be removed, and this patch does so.

Additionally, this patch adds the missing registration of the NexGen CPUs by the cpu_vendor_dev_register() macro.

This patch was build-tested using the defconfig for i386. It also survived a few randconfig instances. Runtime tests were performed by booting an i386 box up to the shell prompt.

---
Peter,

Thanks for your review. Hopefully this version is OK.

 arch/x86/kernel/cpu/nexgen.c |    6 +-----
 1 files changed, 1 insertions(+), 5 deletions(-)
arch/x86/kernel/cpu/nexgen.c +1 -5
index 5d5e1c1..3117e96 100644
@@ -52,8 +52,4 @@ static struct cpu_dev nexgen_cpu_dev __cpuinitdata = {
 	.c_identify	= nexgen_identify,
 };
 
-int __init nexgen_init_cpu(void)
-{
-	cpu_devs[X86_VENDOR_NEXGEN] = &nexgen_cpu_dev;
-	return 0;
-}
+cpu_vendor_dev_register(X86_VENDOR_NEXGEN, &nexgen_cpu_dev);
Signature · 1 line
1.5.3
#5 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

In reply to H. Peter Anvin
Quoted 9 lines · H. Peter Anvin
Quoted 7 lines · Dmitri Vorobiev

Nobody calls the function nexgen_init_cpu(), therefore it can be removed. The variable nexgen_cpu_dev is used only by this function, so this patch removes useless data as well. This patch was build-tested using defconfigs for i386 and x86_64. It also survived a few randconfig instances for i386. Runtime tests were performed by booting 32-bit and 64-bit x86 boxen up to the shell prompt.

Quoted 4 lines

NAK. This is broken.

The reason nexgen_cpu_dev is unreferenced is because the file is missing a cpu_vendor_dev_register(), which should be added.

If we did that long without it wouldn't it be better to just remove it?

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

-Andi

#6 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Andi Kleen пишет:

Quoted 12 lines
Quoted 10 lines · H. Peter Anvin
Quoted 7 lines · Dmitri Vorobiev

Nobody calls the function nexgen_init_cpu(), therefore it can be removed. The variable nexgen_cpu_dev is used only by this function, so this patch removes useless data as well. This patch was build-tested using defconfigs for i386 and x86_64. It also survived a few randconfig instances for i386. Runtime tests were performed by booting 32-bit and 64-bit x86 boxen up to the shell prompt.

NAK. This is broken.

Quoted 9 lines
Quoted 2 lines

The reason nexgen_cpu_dev is unreferenced is because the file is missing a cpu_vendor_dev_register(), which should be added.

If we did that long without it wouldn't it be better to just remove it?

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

Dmitri

Quoted 2 lines

-Andi

#7 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Quoted 8 lines · Dmitri Vorobiev
Quoted 4 lines

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

If it never shipped, then yes, it should be safe to remove. We don't support other vendors that never went to product, e.g. RISE or Montalvo.

-hpa

#8 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

H. Peter Anvin пишет:

Quoted 6 lines
Quoted 5 lines · Dmitri Vorobiev
Quoted 4 lines

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

Quoted 6 lines
Quoted 3 lines

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

If it never shipped, then yes, it should be safe to remove. We don't support other vendors that never went to product, e.g. RISE or Montalvo.

OK, then please drop this patch from the series (the other changes are OK, I presume), and I'll try to take care of NexGen removal with a separate patch.

Thanks,
Dmitri

Quoted 3 lines

-hpa

reviewer Acked-by
#9 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Quoted 4 lines · Dmitri Vorobiev

OK, then please drop this patch from the series (the other changes are OK, I presume), and I'll try to take care of NexGen removal with a separate patch.

... on the rest.

-hpa

#10 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

In reply to H. Peter Anvin
Quoted 11 lines · H. Peter Anvin
Quoted 8 lines · Dmitri Vorobiev
Quoted 4 lines

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

If it never shipped, then yes, it should be safe to remove.

It was shipped as AMD K6, but I have no idea if this file is supposed to handle real k6s or just early nexgen engineering samples.

But since it has been basically broken forever (or at least for all of 2.6) it is probably safe to remove until someone complains.

-Andi

#11 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Quoted 6 lines · Andi Kleen

It was shipped as AMD K6, but I have no idea if this file is supposed to handle real k6s or just early nexgen engineering samples.

But since it has been basically broken forever (or at least for all of 2.6) it is probably safe to remove until someone complains.

The file only triggers on the "NexGenDriven" CPU signature. Let's kill it.

-hpa

#12 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

In reply to H. Peter Anvin
Quoted 6 lines · On Sun, 20 Apr 2008 15:34:53 -0400 "H. Peter Anvin" <[email protected]> wrote:
Quoted 5 lines · Dmitri Vorobiev
Quoted 4 lines

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

Quoted 6 lines
Quoted 3 lines

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

If it never shipped, then yes, it should be safe to remove. We don't support other vendors that never went to product, e.g. RISE or Montalvo.

There are/were a few Nexgen boxes out there - strange beast where the BIOS code emulated some instructions. So few I doubt killing it off is a problem at all.

Alan

#13 lore raw

Re: [PATCH 2/4] x86: remove unused code and data in nexgen.c

Alan Cox пишет:

Quoted 10 lines
Quoted 7 lines · On Sun, 20 Apr 2008 15:34:53 -0400 "H. Peter Anvin" <[email protected]> wrote:
Quoted 6 lines · Dmitri Vorobiev
Quoted 3 lines

Also NexGen was never shipped anyways, but bought by AMD and became the K6. So I assume whatever this file is supposed to do is done by amd.c anyways.

IOW, you claim that now it's safe enough to throw away the NexGen altogether?

Quoted 6 lines
Quoted 2 lines

If it never shipped, then yes, it should be safe to remove. We don't support other vendors that never went to product, e.g. RISE or Montalvo.

There are/were a few Nexgen boxes out there - strange beast where the BIOS code emulated some instructions. So few I doubt killing it off is a problem at all.

The patch that kills the NexGen is already in Ingo's x86 tree, therefore the beast is dead, at least from the kernel viewpoint.

Thanks,
Dmitri

Quoted 3 lines

Alan

#14 lore raw

PATCH 3/4 x86_64: the standard_io_resources[] array can become static

In reply to Dmitri Vorobiev

In arch/x86/kernel/setup_64.c, the standard_io_resources array is needlessly defined as global. This patch makes this variable static.

This patch was successfully build-tested using the defconfig for x86_64. Runtime test was performed by booting a 64-bit x86 box up to the shell prompt.

---
 arch/x86/kernel/setup_64.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/setup_64.c +1 -1
index cd245ef..0332ae2 100644
@@ -117,7 +117,7 @@ extern int root_mountflags;
 
 char __initdata command_line[COMMAND_LINE_SIZE];
 
-struct resource standard_io_resources[] = {
+static struct resource standard_io_resources[] = {
 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 	{ .name = "pic1", .start = 0x20, .end = 0x21,
Signature · 1 line
1.5.3
#15 lore raw

PATCH 4/4 x86: conditionally compile sysfs stuff in intel_cacheinfo.c

In reply to Dmitri Vorobiev

Four functions defined in arch/x86/kernel/cpu/intel_cacheinfo.c:

cache_shared_cpu_map_setup()
cache_remove_shared_cpu_map()
free_cache_attributes()
detect_cache_attributes()

are needed only in the code that is conditionally compiled when CONFIG_SYSFS is defined. The same is true about the cpuid4_info[] array. This patch places the definitions of these functions and the array inside the corresponding ifdef.

This patch was build-tested both with sysfs turned off and on. The kernel was built successfully in both cases. Runtime tests were performed using x86 and x86_64 machines, which booted to the shell prompt successfully.

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/x86/kernel/cpu/intel_cacheinfo.c +2 -2
index 1b88986..d743dfe 100644
@@ -450,6 +450,8 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c)
 	return l2;
 }
 
+#ifdef CONFIG_SYSFS
+
 /* pointer to _cpuid4_info array (for each cache leaf) */
 static struct _cpuid4_info *cpuid4_info[NR_CPUS];
 #define CPUID4_INFO_IDX(x,y)    (&((cpuid4_info[x])[y]))
@@ -553,8 +555,6 @@ out:
 	return retval;
 }
 
-#ifdef CONFIG_SYSFS
-
 #include <linux/kobject.h>
 #include <linux/sysfs.h>
 
Signature · 1 line
1.5.3
#16 lore raw
In reply to Dmitri Vorobiev
Quoted 6 lines · Dmitri Vorobiev

Hi Ingo, Peter, Thomas,

The following patches contain a few clean-ups, mostly stemming from analyzis of sparse messages.

The patches are against current x86.git.

thanks Dmitri, applied #1/#2/#4, as per the discussion.

Ingo

A message that is not in the archive (sent off-list or not yet received): [email protected]

#18 lore raw
Quoted 7 lines · Dmitri Vorobiev
Quoted 1 line

thanks Dmitri, applied #1/#2/#4, as per the discussion.

Hi Ingo,

This should be #1, #3 and #4. The NexGen patch, which was #2 in the original series, have been replaced by a bigger patch that kills NexGen altogether. Plese check, if you have time for that.

yeah, i did the right thing and described the wrong selection :) You can check the end result via:

http://people.redhat.com/mingo/x86.git/README

Ingo

#19 lore raw

Ingo Molnar пишет:

Quoted 5 lines · Dmitri Vorobiev
Quoted 4 lines
Quoted 2 lines

thanks Dmitri, applied #1/#2/#4, as per the discussion.

Hi Ingo,

Quoted 8 lines
Quoted 3 lines

This should be #1, #3 and #4. The NexGen patch, which was #2 in the original series, have been replaced by a bigger patch that kills NexGen altogether. Plese check, if you have time for that.

yeah, i did the right thing and described the wrong selection :) You can check the end result via:

http://people.redhat.com/mingo/x86.git/README

... which indicates that #4 is still not there, which is a pity, since the #4 patch gave me a more-than-negligible binary shrinkage for the case CONFIG_SYSFS=n :)

Dmitri

Quoted 3 lines

Ingo

#20 lore raw
Quoted 4 lines · Dmitri Vorobiev
Quoted 4 lines
Quoted 3 lines

This should be #1, #3 and #4. The NexGen patch, which was #2 in the original series, have been replaced by a bigger patch that kills NexGen altogether. Plese check, if you have time for that.

Quoted 8 lines
Quoted 4 lines

yeah, i did the right thing and described the wrong selection :) You can check the end result via:

http://people.redhat.com/mingo/x86.git/README

... which indicates that #4 is still not there, which is a pity, since the #4 patch gave me a more-than-negligible binary shrinkage for the case CONFIG_SYSFS=n :)

oh, yes, i forgot - that collided with some changes to that file - percpu changes and 4Kcpu support. I think we should delay #4 to a bit later, when the other changes have gone in - or you could try to send a patch against sched-devel.git/latest which should show the merged up status of that code:

http://people.redhat.com/mingo/sched-devel.git/README

sched-devel/latest has all of the x86 tree 'embedded' in it - plus the 4Kcpu changes as well.

Ingo

#21 lore raw

Ingo Molnar пишет:

Quoted 19 lines · Dmitri Vorobiev
Quoted 11 lines
Quoted 8 lines
Quoted 4 lines

This should be #1, #3 and #4. The NexGen patch, which was #2 in the original series, have been replaced by a bigger patch that kills NexGen altogether. Plese check, if you have time for that.

yeah, i did the right thing and described the wrong selection :) You can check the end result via:

http://people.redhat.com/mingo/x86.git/README

... which indicates that #4 is still not there, which is a pity, since the #4 patch gave me a more-than-negligible binary shrinkage for the case CONFIG_SYSFS=n :)

oh, yes, i forgot - that collided with some changes to that file - percpu changes and 4Kcpu support. I think we should delay #4 to a bit later, when the other changes have gone in - or you could try to send a patch against sched-devel.git/latest which should show the merged up status of that code:

Quoted 4 lines

http://people.redhat.com/mingo/sched-devel.git/README

sched-devel/latest has all of the x86 tree 'embedded' in it - plus the 4Kcpu changes as well.

OK, let's wait till the things calm down a bit. The "sysfs disabled" case should be rare enough to not be in a big rush here.

I'll resubmit this after the merge window has closed.

Thanks,
Dmitri

Quoted 3 lines

Ingo