Skip to contentSkip to conversation

staging: zsmalloc: Fix link error on ARM

PATCH In review Joerg Roedel 3 participants 11 messages Mar 26, 2013 – Mar 27, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 staging: zsmalloc: Fix link error on ARM

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly.

---
 drivers/staging/zsmalloc/zsmalloc-main.c |    5 +----
 mm/vmalloc.c                             |    1 +
 2 files changed, 2 insertions(+), 4 deletions(-)
drivers/staging/zsmalloc/zsmalloc-main.c +1 -4
index e78d262..324e123 100644
@@ -656,11 +656,8 @@ static inline void __zs_unmap_object(struct mapping_area *area,
 				struct page *pages[2], int off, int size)
 {
 	unsigned long addr = (unsigned long)area->vm_addr;
-	unsigned long end = addr + (PAGE_SIZE * 2);
 
-	flush_cache_vunmap(addr, end);
-	unmap_kernel_range_noflush(addr, PAGE_SIZE * 2);
-	flush_tlb_kernel_range(addr, end);
+	unmap_kernel_range(addr, PAGE_SIZE * 2);
 }
 
 #else /* USE_PGTABLE_MAPPING */
mm/vmalloc.c +1 -0
index 0f751f2..f7cba11 100644
@@ -1266,6 +1266,7 @@ void unmap_kernel_range(unsigned long addr, unsigned long size)
 	vunmap_page_range(addr, end);
 	flush_tlb_kernel_range(addr, end);
 }
+EXPORT_SYMBOL_GPL(unmap_kernel_range);
 
 int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages)
 {
Signature · 8 lines
1.7.9.5


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#1 lore raw
Quoted 8 lines · Joerg Roedel

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly.

Why is this not an error for any other architecture? Why is arm special?

Quoted 27 lines
---
 drivers/staging/zsmalloc/zsmalloc-main.c |    5 +----
 mm/vmalloc.c                             |    1 +
 2 files changed, 2 insertions(+), 4 deletions(-)
drivers/staging/zsmalloc/zsmalloc-main.c +1 -4
index e78d262..324e123 100644
@@ -656,11 +656,8 @@ static inline void __zs_unmap_object(struct mapping_area *area,
 				struct page *pages[2], int off, int size)
 {
 	unsigned long addr = (unsigned long)area->vm_addr;
-	unsigned long end = addr + (PAGE_SIZE * 2);
 
-	flush_cache_vunmap(addr, end);
-	unmap_kernel_range_noflush(addr, PAGE_SIZE * 2);
-	flush_tlb_kernel_range(addr, end);
+	unmap_kernel_range(addr, PAGE_SIZE * 2);
 }
 
 #else /* USE_PGTABLE_MAPPING */
mm/vmalloc.c +0 -0
index 0f751f2..f7cba11 100644

mm/vmalloc.c void unmap_kernel_range(unsigned long addr, unsigned long size) · line 1269

Quoted 4 lines
mm/vmalloc.c +1 -0
@@ -1266,6 +1266,7 @@ void unmap_kernel_range(unsigned long addr, unsigned long size)
 	vunmap_page_range(addr, end);
 	flush_tlb_kernel_range(addr, end);
 }
+EXPORT_SYMBOL_GPL(unmap_kernel_range);

I _really_ don't like adding core exports for a staging driver, there is no guarantee that the staging driver will not just be deleted tomorrow.

So, if at all possible, I don't want to do this.

Perhaps, just make it so the staging code can't be a module?

greg k-h

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#2 lore raw
Quoted 8 lines · Greg Kroah-Hartman
Quoted 7 lines · Joerg Roedel

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly.

Quoted 4 lines
Quoted 1 line

Why is this not an error for any other architecture? Why is arm special?

The version of the function __zs_unmap_object() which uses flush_tlb_kernel_range() in the zsmalloc driver is only compiled in when USE_PGTABLE_MAPPING is defined. And USE_PGTABLE_MAPPING is defined in the same file only when CONFIG_ARM is defined. So this happens only on ARM.

Regards,

Joerg

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#3 lore raw
Quoted 11 lines · Joerg Roedel
Quoted 10 lines · Greg Kroah-Hartman
Quoted 8 lines · Joerg Roedel

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly.

Quoted 8 lines
Quoted 2 lines

Why is this not an error for any other architecture? Why is arm special?

The version of the function __zs_unmap_object() which uses flush_tlb_kernel_range() in the zsmalloc driver is only compiled in when USE_PGTABLE_MAPPING is defined. And USE_PGTABLE_MAPPING is defined in the same file only when CONFIG_ARM is defined. So this happens only on ARM.

Then should I just mark the driver as broken on ARM?

Any reason for not including the driver authors on the Cc: for this patch?

greg k-h

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#4 lore raw
Quoted 3 lines · Greg Kroah-Hartman

On Wed, Mar 27, 2013 at 12:03:59AM +0100, Joerg Roedel wrote:

Then should I just mark the driver as broken on ARM?

Well, at least ARM on SMP, for !SMP the missing function is defined and will be inlined.

Quoted 2 lines

Any reason for not including the driver authors on the Cc: for this patch?

I wasn't sure who the driver author is in the long list of persons from the get_maintainer script. So it was more or less lazyness :)

Should I try to get that patch through the driver authors instead of you?

Joerg

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#5 lore raw
Quoted 7 lines · Joerg Roedel
Quoted 3 lines · Greg Kroah-Hartman

On Wed, Mar 27, 2013 at 12:03:59AM +0100, Joerg Roedel wrote:

Then should I just mark the driver as broken on ARM?

Well, at least ARM on SMP, for !SMP the missing function is defined and will be inlined.

Then some Kconfig dependancies could be tweaked to try to make this work properly, not building the driver where it will be broken.

Quoted 8 lines
Quoted 2 lines

Any reason for not including the driver authors on the Cc: for this patch?

I wasn't sure who the driver author is in the long list of persons from the get_maintainer script. So it was more or less lazyness :)

Should I try to get that patch through the driver authors instead of you?

At least cc: them to have them weigh in on the issue, I know they are trying to do something in this area, but I didn't think it was the same as your patch.

thanks,

greg k-h

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#7 lore raw
In reply to Joerg Roedel
Quoted 8 lines · Joerg Roedel

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly.

Oops, it was my fault. When I tested [1] on CONFIG_SMP machine on ARM, it worked well. It means it's not always problem on every CONFIG_SMP on ARM machine but some SMP machine define flush_tlb_kernel_range, others don't.

At that time, Russell King already suggested same thing with your patch and I meant to clean it up because the patch was already merged but I didn't. Because we didn't catch up that it breaks build on some configuration so I thought it's just clean up patch and Greg didn't want to accept NOT-BUG patch of any z* family.

Now, it's BUG patch.

Remained problem is that Greg doesn't want to export core function for staging driver and it's reasonable for me.
So my opinion is remove zsmalloc module build and could recover it with making unmap_kernel_range exported function after we merged it into mainline.

And please Cc stable.

[1] [99155188, zsmalloc: Fix TLB coherency and build problem]

Quoted 43 lines
---
 drivers/staging/zsmalloc/zsmalloc-main.c |    5 +----
 mm/vmalloc.c                             |    1 +
 2 files changed, 2 insertions(+), 4 deletions(-)
drivers/staging/zsmalloc/zsmalloc-main.c +1 -4
index e78d262..324e123 100644
@@ -656,11 +656,8 @@ static inline void __zs_unmap_object(struct mapping_area *area,
 				struct page *pages[2], int off, int size)
 {
 	unsigned long addr = (unsigned long)area->vm_addr;
-	unsigned long end = addr + (PAGE_SIZE * 2);
 
-	flush_cache_vunmap(addr, end);
-	unmap_kernel_range_noflush(addr, PAGE_SIZE * 2);
-	flush_tlb_kernel_range(addr, end);
+	unmap_kernel_range(addr, PAGE_SIZE * 2);
 }
 
 #else /* USE_PGTABLE_MAPPING */
mm/vmalloc.c +1 -0
index 0f751f2..f7cba11 100644
@@ -1266,6 +1266,7 @@ void unmap_kernel_range(unsigned long addr, unsigned long size)
 	vunmap_page_range(addr, end);
 	flush_tlb_kernel_range(addr, end);
 }
+EXPORT_SYMBOL_GPL(unmap_kernel_range);
 
 int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages)
 {
Signature · 8 lines
1.7.9.5


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Signature · 8 lines
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#8 lore raw
Quoted 11 lines · Minchan Kim

Oops, it was my fault. When I tested [1] on CONFIG_SMP machine on ARM, it worked well. It means it's not always problem on every CONFIG_SMP on ARM machine but some SMP machine define flush_tlb_kernel_range, others don't.

At that time, Russell King already suggested same thing with your patch and I meant to clean it up because the patch was already merged but I didn't. Because we didn't catch up that it breaks build on some configuration so I thought it's just clean up patch and Greg didn't want to accept NOT-BUG patch of any z* family.

Quoted 4 lines

Now, it's BUG patch.

Remained problem is that Greg doesn't want to export core function for staging driver and it's reasonable for me.

Okay, I see. So that is probably also the reason for the reimplementation of unmap_kernel_range in the zsmalloc module :)

Quoted 3 lines

So my opinion is remove zsmalloc module build and could recover it with making unmap_kernel_range exported function after we merged it into mainline.

Sounds reasonable, I update the patch to only allow zsmalloc to be built-in. The benefit is that this still allows to use unmap_kernel_range() in the driver.

Thanks,

Joerg

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#9 lore raw
In reply to Minchan Kim
Quoted 1 line · Minchan Kim

And please Cc stable.

Okay, here it is. The result is compile-tested.

Changes since v1:

* Remove the module-export for unmap_kernel_range and make zsmalloc built-in instead

Here is the patch:

Acked-by
#10 lore raw
Quoted 26 lines · Joerg Roedel
Quoted 1 line · Minchan Kim

And please Cc stable.

Okay, here it is. The result is compile-tested.

Changes since v1:

* Remove the module-export for unmap_kernel_range and make zsmalloc built-in instead

Here is the patch:

Quoted 1 line

From 2b70502720b36909f9f39bdf27be21321a219c31 Mon Sep 17 00:00:00 2001

From: Joerg Roedel <[email protected]>
Date: Tue, 26 Mar 2013 23:24:22 +0100
Subject: [PATCH v2] staging: zsmalloc: Fix link error on ARM

Testing the arm chromebook config against the upstream kernel produces a linker error for the zsmalloc module from staging. The symbol flush_tlb_kernel_range is not available there. Fix this by removing the reimplementation of unmap_kernel_range in the zsmalloc module and using the function directly. The unmap_kernel_range function is not usable by modules, so also disallow building the driver as a module for now.

Quoted 2 lines
Signature · 8 lines
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>