Re: linux-next: build failure after merge of the mm tree

Re: linux-next: build failure after merge of the mm tree

[Matthew Wilcox]

On Tue, Aug 22, 2023 at 09:55:37AM +1000, Stephen Rothwell wrote:
> In file included from include/trace/trace_events.h:27,
>                  from include/trace/define_trace.h:102,
>                  from fs/xfs/xfs_trace.h:4428,
>                  from fs/xfs/xfs_trace.c:45:
> include/linux/pgtable.h:8:25: error: initializer element is not constant
>     8 | #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)

Ummm.  PowerPC doesn't have a compile-time constant PMD size?

arch/powerpc/include/asm/book3s/64/pgtable.h:#define PMD_SHIFT  (PAGE_SHIFT + PTE_INDEX_SIZE)
arch/powerpc/include/asm/book3s/64/pgtable.h:#define PTE_INDEX_SIZE  __pte_index_size

That's really annoying.  I'll try to work around it.

Re: linux-next: build failure after merge of the mm tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 711 bytes --]

Hi Matthew,

On Tue, 22 Aug 2023 02:11:44 +0100 Matthew Wilcox <willy@infradead.org> wrote:
>
> On Tue, Aug 22, 2023 at 09:55:37AM +1000, Stephen Rothwell wrote:
> > In file included from include/trace/trace_events.h:27,
> >                  from include/trace/define_trace.h:102,
> >                  from fs/xfs/xfs_trace.h:4428,
> >                  from fs/xfs/xfs_trace.c:45:
> > include/linux/pgtable.h:8:25: error: initializer element is not constant
> >     8 | #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)  
> 
> Ummm.  PowerPC doesn't have a compile-time constant PMD size?

Yeah, you are not the first (or probably the last) to be caught by that.

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the mm tree

[Matthew Wilcox]

On Tue, Aug 22, 2023 at 11:22:17AM +1000, Stephen Rothwell wrote:
> Hi Matthew,
> 
> On Tue, 22 Aug 2023 02:11:44 +0100 Matthew Wilcox <willy@infradead.org> wrote:
> >
> > On Tue, Aug 22, 2023 at 09:55:37AM +1000, Stephen Rothwell wrote:
> > > In file included from include/trace/trace_events.h:27,
> > >                  from include/trace/define_trace.h:102,
> > >                  from fs/xfs/xfs_trace.h:4428,
> > >                  from fs/xfs/xfs_trace.c:45:
> > > include/linux/pgtable.h:8:25: error: initializer element is not constant
> > >     8 | #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)  
> > 
> > Ummm.  PowerPC doesn't have a compile-time constant PMD size?
> 
> Yeah, you are not the first (or probably the last) to be caught by that.

I think this will do the trick.  Any comments?

diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 1904eaf7a2e9..d5a4e6c2dcd1 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -796,15 +796,6 @@ DEFINE_INODE_EVENT(xfs_inode_reclaiming);
 DEFINE_INODE_EVENT(xfs_inode_set_need_inactive);
 DEFINE_INODE_EVENT(xfs_inode_inactivating);
 
-/*
- * ftrace's __print_symbolic requires that all enum values be wrapped in the
- * TRACE_DEFINE_ENUM macro so that the enum value can be encoded in the ftrace
- * ring buffer.  Somehow this was only worth mentioning in the ftrace sample
- * code.
- */
-TRACE_DEFINE_ENUM(PMD_ORDER);
-TRACE_DEFINE_ENUM(PUD_ORDER);
-
 TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_SHARED);
 TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_COW);
 
@@ -823,13 +814,10 @@ TRACE_EVENT(xfs_filemap_fault,
 		__entry->order = order;
 		__entry->write_fault = write_fault;
 	),
-	TP_printk("dev %d:%d ino 0x%llx %s write_fault %d",
+	TP_printk("dev %d:%d ino 0x%llx order:%u write_fault %d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->ino,
-		  __print_symbolic(__entry->order,
-			{ 0,		"PTE" },
-			{ PMD_ORDER,	"PMD" },
-			{ PUD_ORDER,	"PUD" }),
+		  __entry->order,
 		  __entry->write_fault)
 )
 

Re: linux-next: build failure after merge of the mm tree

[Darrick J. Wong]

On Tue, Aug 22, 2023 at 02:34:06AM +0100, Matthew Wilcox wrote:
> On Tue, Aug 22, 2023 at 11:22:17AM +1000, Stephen Rothwell wrote:
> > Hi Matthew,
> > 
> > On Tue, 22 Aug 2023 02:11:44 +0100 Matthew Wilcox <willy@infradead.org> wrote:
> > >
> > > On Tue, Aug 22, 2023 at 09:55:37AM +1000, Stephen Rothwell wrote:
> > > > In file included from include/trace/trace_events.h:27,
> > > >                  from include/trace/define_trace.h:102,
> > > >                  from fs/xfs/xfs_trace.h:4428,
> > > >                  from fs/xfs/xfs_trace.c:45:
> > > > include/linux/pgtable.h:8:25: error: initializer element is not constant
> > > >     8 | #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)  
> > > 
> > > Ummm.  PowerPC doesn't have a compile-time constant PMD size?
> > 
> > Yeah, you are not the first (or probably the last) to be caught by that.
> 
> I think this will do the trick.  Any comments?
> 
> diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
> index 1904eaf7a2e9..d5a4e6c2dcd1 100644
> --- a/fs/xfs/xfs_trace.h
> +++ b/fs/xfs/xfs_trace.h
> @@ -796,15 +796,6 @@ DEFINE_INODE_EVENT(xfs_inode_reclaiming);
>  DEFINE_INODE_EVENT(xfs_inode_set_need_inactive);
>  DEFINE_INODE_EVENT(xfs_inode_inactivating);
>  
> -/*
> - * ftrace's __print_symbolic requires that all enum values be wrapped in the
> - * TRACE_DEFINE_ENUM macro so that the enum value can be encoded in the ftrace
> - * ring buffer.  Somehow this was only worth mentioning in the ftrace sample
> - * code.
> - */

Please leave this ^^^ comment, because the need for TRACE_DEFINE_ENUM to
make enums work in tracepoints is not at all obvious.

> -TRACE_DEFINE_ENUM(PMD_ORDER);
> -TRACE_DEFINE_ENUM(PUD_ORDER);
> -
>  TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_SHARED);
>  TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_COW);
>  
> @@ -823,13 +814,10 @@ TRACE_EVENT(xfs_filemap_fault,
>  		__entry->order = order;
>  		__entry->write_fault = write_fault;
>  	),
> -	TP_printk("dev %d:%d ino 0x%llx %s write_fault %d",
> +	TP_printk("dev %d:%d ino 0x%llx order:%u write_fault %d",

"order %u" to match the (non dev_t) style of the rest of the xfs
tracepoints.

--D

>  		  MAJOR(__entry->dev), MINOR(__entry->dev),
>  		  __entry->ino,
> -		  __print_symbolic(__entry->order,
> -			{ 0,		"PTE" },
> -			{ PMD_ORDER,	"PMD" },
> -			{ PUD_ORDER,	"PUD" }),
> +		  __entry->order,
>  		  __entry->write_fault)
>  )
>  
> 
> 

Re: linux-next: build failure after merge of the mm tree

[Matthew Wilcox]

On Mon, Aug 21, 2023 at 09:00:43PM -0700, Darrick J. Wong wrote:
> Please leave this ^^^ comment, because the need for TRACE_DEFINE_ENUM to
> make enums work in tracepoints is not at all obvious.
> 
> "order %u" to match the (non dev_t) style of the rest of the xfs
> tracepoints.

ACK, thanks.

Andrew, please add this -fix patch for "mm: Remove enum
page_entry_size".

diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 1904eaf7a2e9..fd789e00dfd6 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -802,9 +802,6 @@ DEFINE_INODE_EVENT(xfs_inode_inactivating);
  * ring buffer.  Somehow this was only worth mentioning in the ftrace sample
  * code.
  */
-TRACE_DEFINE_ENUM(PMD_ORDER);
-TRACE_DEFINE_ENUM(PUD_ORDER);
-
 TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_SHARED);
 TRACE_DEFINE_ENUM(XFS_REFC_DOMAIN_COW);
 
@@ -823,13 +820,10 @@ TRACE_EVENT(xfs_filemap_fault,
 		__entry->order = order;
 		__entry->write_fault = write_fault;
 	),
-	TP_printk("dev %d:%d ino 0x%llx %s write_fault %d",
+	TP_printk("dev %d:%d ino 0x%llx order %u write_fault %d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->ino,
-		  __print_symbolic(__entry->order,
-			{ 0,		"PTE" },
-			{ PMD_ORDER,	"PMD" },
-			{ PUD_ORDER,	"PUD" }),
+		  __entry->order,
 		  __entry->write_fault)
 )
 

Re: linux-next: build failure after merge of the mm tree

[Michael Ellerman]

Matthew Wilcox <willy@infradead.org> writes:
> On Tue, Aug 22, 2023 at 09:55:37AM +1000, Stephen Rothwell wrote:
>> In file included from include/trace/trace_events.h:27,
>>                  from include/trace/define_trace.h:102,
>>                  from fs/xfs/xfs_trace.h:4428,
>>                  from fs/xfs/xfs_trace.c:45:
>> include/linux/pgtable.h:8:25: error: initializer element is not constant
>>     8 | #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)
>
> Ummm.  PowerPC doesn't have a compile-time constant PMD size?

Yeah. The joys of supporting two MMUs with different supported page
sizes in a single kernel binary.

> arch/powerpc/include/asm/book3s/64/pgtable.h:#define PMD_SHIFT  (PAGE_SHIFT + PTE_INDEX_SIZE)
> arch/powerpc/include/asm/book3s/64/pgtable.h:#define PTE_INDEX_SIZE  __pte_index_size
>
> That's really annoying.  I'll try to work around it.

Sorry, thanks.

cheers

Re: linux-next: build failure after merge of the mm tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1985 bytes --]

Hi all,

Just cc'ing the PowerPC guys to see if my fix is sensible.

On Mon, 27 Nov 2023 13:28:09 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> After merging the mm tree, today's linux-next build (powerpc64
> allnoconfig) failed like this:
> 
> arch/powerpc/mm/book3s64/pgtable.c:557:5: error: no previous prototype for 'pmd_move_must_withdraw' [-Werror=missing-prototypes]
>   557 | int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
>       |     ^~~~~~~~~~~~~~~~~~~~~~
> cc1: all warnings being treated as errors
> 
> Caused by commit
> 
>   c6345dfa6e3e ("Makefile.extrawarn: turn on missing-prototypes globally")
> 
> I have added the following patch for today (which could be applied to
> the mm or powerpc trees):
> 
> From 194805b44c11b4c0aa28bdcdc0bb0d82acef394c Mon Sep 17 00:00:00 2001
> From: Stephen Rothwell <sfr@canb.auug.org.au>
> Date: Mon, 27 Nov 2023 13:08:57 +1100
> Subject: [PATCH] powerpc: pmd_move_must_withdraw() is only needed for
>  CONFIG_TRANSPARENT_HUGEPAGE
> 
> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
> ---
>  arch/powerpc/mm/book3s64/pgtable.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
> index be229290a6a7..3438ab72c346 100644
> --- a/arch/powerpc/mm/book3s64/pgtable.c
> +++ b/arch/powerpc/mm/book3s64/pgtable.c
> @@ -542,6 +542,7 @@ void ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
>  	set_pte_at(vma->vm_mm, addr, ptep, pte);
>  }
>  
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>  /*
>   * For hash translation mode, we use the deposited table to store hash slot
>   * information and they are stored at PTRS_PER_PMD offset from related pmd
> @@ -563,6 +564,7 @@ int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
>  
>  	return true;
>  }
> +#endif
>  
>  /*
>   * Does the CPU support tlbie?
> -- 
> 2.40.1

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the mm tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 2313 bytes --]

Hi all,

On Mon, 27 Nov 2023 14:48:52 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> 
> Just cc'ing the PowerPC guys to see if my fix is sensible.
> 
> On Mon, 27 Nov 2023 13:28:09 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> >
> > After merging the mm tree, today's linux-next build (powerpc64
> > allnoconfig) failed like this:
> > 
> > arch/powerpc/mm/book3s64/pgtable.c:557:5: error: no previous prototype for 'pmd_move_must_withdraw' [-Werror=missing-prototypes]
> >   557 | int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
> >       |     ^~~~~~~~~~~~~~~~~~~~~~
> > cc1: all warnings being treated as errors
> > 
> > Caused by commit
> > 
> >   c6345dfa6e3e ("Makefile.extrawarn: turn on missing-prototypes globally")
> > 
> > I have added the following patch for today (which could be applied to
> > the mm or powerpc trees):
> > 
> > From 194805b44c11b4c0aa28bdcdc0bb0d82acef394c Mon Sep 17 00:00:00 2001
> > From: Stephen Rothwell <sfr@canb.auug.org.au>
> > Date: Mon, 27 Nov 2023 13:08:57 +1100
> > Subject: [PATCH] powerpc: pmd_move_must_withdraw() is only needed for
> >  CONFIG_TRANSPARENT_HUGEPAGE
> > 
> > Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
> > ---
> >  arch/powerpc/mm/book3s64/pgtable.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
> > index be229290a6a7..3438ab72c346 100644
> > --- a/arch/powerpc/mm/book3s64/pgtable.c
> > +++ b/arch/powerpc/mm/book3s64/pgtable.c
> > @@ -542,6 +542,7 @@ void ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
> >  	set_pte_at(vma->vm_mm, addr, ptep, pte);
> >  }
> >  
> > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> >  /*
> >   * For hash translation mode, we use the deposited table to store hash slot
> >   * information and they are stored at PTRS_PER_PMD offset from related pmd
> > @@ -563,6 +564,7 @@ int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
> >  
> >  	return true;
> >  }
> > +#endif
> >  
> >  /*
> >   * Does the CPU support tlbie?
> > -- 
> > 2.40.1  

I am still carrying this patch (it should probably go into the mm
tree).  Is someone going to pick it up (assuming it is correct)?

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the mm tree

[Andrew Morton]

On Fri, 1 Dec 2023 09:04:39 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:

> Hi all,
> 
> > > diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
> > > index be229290a6a7..3438ab72c346 100644
> > > --- a/arch/powerpc/mm/book3s64/pgtable.c
> > > +++ b/arch/powerpc/mm/book3s64/pgtable.c
> > > @@ -542,6 +542,7 @@ void ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
> > >  	set_pte_at(vma->vm_mm, addr, ptep, pte);
> > >  }
> > >  
> > > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > >  /*
> > >   * For hash translation mode, we use the deposited table to store hash slot
> > >   * information and they are stored at PTRS_PER_PMD offset from related pmd
> > > @@ -563,6 +564,7 @@ int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
> > >  
> > >  	return true;
> > >  }
> > > +#endif
> > >  
> > >  /*
> > >   * Does the CPU support tlbie?
> > > -- 
> > > 2.40.1  
> 
> I am still carrying this patch (it should probably go into the mm
> tree).  Is someone going to pick it up (assuming it is correct)?

AFAIK we're still awaiting input from the ppc team.

I'll grab it.  If it breaks things then we-told-you-so!

Re: linux-next: build failure after merge of the mm tree

[Michael Ellerman]

Stephen Rothwell <sfr@canb.auug.org.au> writes:
> On Mon, 27 Nov 2023 14:48:52 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>> 
>> Just cc'ing the PowerPC guys to see if my fix is sensible.
>> 
>> On Mon, 27 Nov 2023 13:28:09 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>> >
>> > After merging the mm tree, today's linux-next build (powerpc64
>> > allnoconfig) failed like this:
>> > 
>> > arch/powerpc/mm/book3s64/pgtable.c:557:5: error: no previous prototype for 'pmd_move_must_withdraw' [-Werror=missing-prototypes]
>> >   557 | int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
>> >       |     ^~~~~~~~~~~~~~~~~~~~~~
>> > cc1: all warnings being treated as errors
>> > 
>> > Caused by commit
>> > 
>> >   c6345dfa6e3e ("Makefile.extrawarn: turn on missing-prototypes globally")
>> > 
>> > I have added the following patch for today (which could be applied to
>> > the mm or powerpc trees):
>> > 
>> > From 194805b44c11b4c0aa28bdcdc0bb0d82acef394c Mon Sep 17 00:00:00 2001
>> > From: Stephen Rothwell <sfr@canb.auug.org.au>
>> > Date: Mon, 27 Nov 2023 13:08:57 +1100
>> > Subject: [PATCH] powerpc: pmd_move_must_withdraw() is only needed for
>> >  CONFIG_TRANSPARENT_HUGEPAGE
>> > 
>> > Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
>> > ---
>> >  arch/powerpc/mm/book3s64/pgtable.c | 2 ++
>> >  1 file changed, 2 insertions(+)
>> > 
>> > diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
>> > index be229290a6a7..3438ab72c346 100644
>> > --- a/arch/powerpc/mm/book3s64/pgtable.c
>> > +++ b/arch/powerpc/mm/book3s64/pgtable.c
>> > @@ -542,6 +542,7 @@ void ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
>> >  	set_pte_at(vma->vm_mm, addr, ptep, pte);
>> >  }
>> >  
>> > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>> >  /*
>> >   * For hash translation mode, we use the deposited table to store hash slot
>> >   * information and they are stored at PTRS_PER_PMD offset from related pmd
>> > @@ -563,6 +564,7 @@ int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
>> >  
>> >  	return true;
>> >  }
>> > +#endif
>> >  
>> >  /*
>> >   * Does the CPU support tlbie?
>> > -- 
>> > 2.40.1  
>
> I am still carrying this patch (it should probably go into the mm
> tree).  Is someone going to pick it up (assuming it is correct)?

I applied it to my next a few days ago, but I must have forgotten to
push. It's in there now.

cheers

Re: linux-next: build failure after merge of the mm tree

[Andrew Morton]

On Fri, 01 Dec 2023 09:39:20 +1100 Michael Ellerman <mpe@ellerman.id.au> wrote:

> > I am still carrying this patch (it should probably go into the mm
> > tree).  Is someone going to pick it up (assuming it is correct)?
> 
> I applied it to my next a few days ago, but I must have forgotten to
> push. It's in there now.

I'll keep a copy in mm.git, to keep the dependencies nice.  I added
your acked-by.

Re: linux-next: build failure after merge of the mm tree

[Michael Ellerman]

Andrew Morton <akpm@linux-foundation.org> writes:
> On Fri, 01 Dec 2023 09:39:20 +1100 Michael Ellerman <mpe@ellerman.id.au> wrote:
>
>> > I am still carrying this patch (it should probably go into the mm
>> > tree).  Is someone going to pick it up (assuming it is correct)?
>> 
>> I applied it to my next a few days ago, but I must have forgotten to
>> push. It's in there now.
>
> I'll keep a copy in mm.git, to keep the dependencies nice.  I added
> your acked-by.

Sure thing. Thanks.

cheers

linux-next: build failure after merge of the mm tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 2280 bytes --]

Hi all,

After merging the mm tree, today's linux-next build (powerpc
ppc44x_defconfig) failed like this:

In file included from arch/powerpc/include/asm/page.h:247,
                 from arch/powerpc/include/asm/thread_info.h:13,
                 from include/linux/thread_info.h:60,
                 from include/asm-generic/preempt.h:5,
                 from ./arch/powerpc/include/generated/asm/preempt.h:1,
                 from include/linux/preempt.h:78,
                 from include/linux/spinlock.h:56,
                 from include/linux/ipc.h:5,
                 from include/uapi/linux/sem.h:5,
                 from include/linux/sem.h:5,
                 from include/linux/compat.h:14,
                 from arch/powerpc/kernel/asm-offsets.c:12:
arch/powerpc/include/asm/page_32.h:16: warning: "ARCH_DMA_MINALIGN" redefined
   16 | #define ARCH_DMA_MINALIGN       L1_CACHE_BYTES
      | 
In file included from include/linux/time.h:5,
                 from include/linux/compat.h:10:
include/linux/cache.h:104: note: this is the location of the previous definition
  104 | #define ARCH_DMA_MINALIGN __alignof__(unsigned long long)
      | 

(lots of theses)

Caused by commit

  cc7335787e73 ("mm/slab: decouple ARCH_KMALLOC_MINALIGN from ARCH_DMA_MINALIGN")

I have applied the following hack for today - we need something better.

From: Stephen Rothwell <sfr@canb.auug.org.au>
Date: Tue, 13 Jun 2023 16:07:16 +1000
Subject: [PATCH] fix up for "mm/slab: decouple ARCH_KMALLOC_MINALIGN from ARCH_DMA_MINALIGN"

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 arch/powerpc/include/asm/cache.h | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/arch/powerpc/include/asm/cache.h b/arch/powerpc/include/asm/cache.h
index ae0a68a838e8..e9be1396dfd1 100644
--- a/arch/powerpc/include/asm/cache.h
+++ b/arch/powerpc/include/asm/cache.h
@@ -142,5 +142,14 @@ static inline void iccci(void *addr)
 }
 
 #endif /* !__ASSEMBLY__ */
+
+#ifndef __powerpc64__
+#ifdef CONFIG_NOT_COHERENT_CACHE
+#ifndef ARCH_DMA_MINALIGN
+#define ARCH_DMA_MINALIGN	L1_CACHE_BYTES
+#endif
+#endif
+#endif
+
 #endif /* __KERNEL__ */
 #endif /* _ASM_POWERPC_CACHE_H */
-- 
2.39.2

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the mm tree

[Catalin Marinas]

Hi Stephen,

On Tue, Jun 13, 2023 at 04:21:19PM +1000, Stephen Rothwell wrote:
> After merging the mm tree, today's linux-next build (powerpc
> ppc44x_defconfig) failed like this:
> 
> In file included from arch/powerpc/include/asm/page.h:247,
>                  from arch/powerpc/include/asm/thread_info.h:13,
>                  from include/linux/thread_info.h:60,
>                  from include/asm-generic/preempt.h:5,
>                  from ./arch/powerpc/include/generated/asm/preempt.h:1,
>                  from include/linux/preempt.h:78,
>                  from include/linux/spinlock.h:56,
>                  from include/linux/ipc.h:5,
>                  from include/uapi/linux/sem.h:5,
>                  from include/linux/sem.h:5,
>                  from include/linux/compat.h:14,
>                  from arch/powerpc/kernel/asm-offsets.c:12:
> arch/powerpc/include/asm/page_32.h:16: warning: "ARCH_DMA_MINALIGN" redefined
>    16 | #define ARCH_DMA_MINALIGN       L1_CACHE_BYTES
>       | 
> In file included from include/linux/time.h:5,
>                  from include/linux/compat.h:10:
> include/linux/cache.h:104: note: this is the location of the previous definition
>   104 | #define ARCH_DMA_MINALIGN __alignof__(unsigned long long)
>       | 
> 
> (lots of theses)
> 
> Caused by commit
> 
>   cc7335787e73 ("mm/slab: decouple ARCH_KMALLOC_MINALIGN from ARCH_DMA_MINALIGN")
> 
> I have applied the following hack for today - we need something better.

I just posted this series fixing it for powerpc, microblaze and sh. I
did not add the #ifndef __powerpc64__ line since
CONFIG_NOT_COHERENT_CACHE should not be enabled for those builds.

https://lore.kernel.org/r/20230613155245.1228274-1-catalin.marinas@arm.com

> From: Stephen Rothwell <sfr@canb.auug.org.au>
> Date: Tue, 13 Jun 2023 16:07:16 +1000
> Subject: [PATCH] fix up for "mm/slab: decouple ARCH_KMALLOC_MINALIGN from ARCH_DMA_MINALIGN"
> 
> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
> ---
>  arch/powerpc/include/asm/cache.h | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/arch/powerpc/include/asm/cache.h b/arch/powerpc/include/asm/cache.h
> index ae0a68a838e8..e9be1396dfd1 100644
> --- a/arch/powerpc/include/asm/cache.h
> +++ b/arch/powerpc/include/asm/cache.h
> @@ -142,5 +142,14 @@ static inline void iccci(void *addr)
>  }
>  
>  #endif /* !__ASSEMBLY__ */
> +
> +#ifndef __powerpc64__
> +#ifdef CONFIG_NOT_COHERENT_CACHE
> +#ifndef ARCH_DMA_MINALIGN
> +#define ARCH_DMA_MINALIGN	L1_CACHE_BYTES
> +#endif
> +#endif
> +#endif
> +
>  #endif /* __KERNEL__ */
>  #endif /* _ASM_POWERPC_CACHE_H */

I think it should also remove the ARCH_DMA_MINALIGN from asm/page.h (as
I did in my series; sorry I did not cc you, only noticed now that you
reported it as well).

-- 
Catalin