linux-next: manual merge of the sparseirq tree

linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[Ingo, thanks for the heads up about this ...]

Hi Ingo,

Today's linux-next merge of the sparseirq tree got a conflict in
arch/x86/xen/spinlock.c between commit
168d2f464ab9860f0d1e66cf1f9684973222f1c6 ("xen: save previous spinlock
when blocking") from the x86 tree and commit
fb6dc57946f9ebfeac546dd0698d9f065c191668 ("x86: move kstat_irqs from
kstat to irq_desc") from the sparseirq tree.

I fixed it up (see below) and can carry it.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

diff --cc arch/x86/xen/spinlock.c
index d072823,5a48aba..0000000
--- a/arch/x86/xen/spinlock.c
+++ b/arch/x86/xen/spinlock.c
@@@ -194,60 -73,24 +194,60 @@@ static noinline int xen_spin_lock_slow(
  	if (irq == -1)
  		return 0;
  
 +	start = spin_time_start();
 +
  	/* announce we're spinning */
 -	spinning_lock(xl);
 +	prev = spinning_lock(xl);
 +
 +	flags = __raw_local_save_flags();
 +	if (irq_enable) {
 +		ADD_STATS(taken_slow_irqenable, 1);
 +		raw_local_irq_enable();
 +	}
 +
 +	ADD_STATS(taken_slow, 1);
 +	ADD_STATS(taken_slow_nested, prev != NULL);
  
 -	/* clear pending */
 -	xen_clear_irq_pending(irq);
 +	do {
 +		/* clear pending */
 +		xen_clear_irq_pending(irq);
 +
 +		/* check again make sure it didn't become free while
 +		   we weren't looking  */
 +		ret = xen_spin_trylock(lock);
 +		if (ret) {
 +			ADD_STATS(taken_slow_pickup, 1);
 +
 +			/*
 +			 * If we interrupted another spinlock while it
 +			 * was blocking, make sure it doesn't block
 +			 * without rechecking the lock.
 +			 */
 +			if (prev != NULL)
 +				xen_set_irq_pending(irq);
 +			goto out;
 +		}
  
 -	/* check again make sure it didn't become free while
 -	   we weren't looking  */
 -	ret = xen_spin_trylock(lock);
 -	if (ret)
 -		goto out;
 +		/*
 +		 * Block until irq becomes pending.  If we're
 +		 * interrupted at this point (after the trylock but
 +		 * before entering the block), then the nested lock
 +		 * handler guarantees that the irq will be left
 +		 * pending if there's any chance the lock became free;
 +		 * xen_poll_irq() returns immediately if the irq is
 +		 * pending.
 +		 */
 +		xen_poll_irq(irq);
 +		ADD_STATS(taken_slow_spurious, !xen_test_irq_pending(irq));
 +	} while (!xen_test_irq_pending(irq)); /* check for spurious wakeups */
  
- 	kstat_this_cpu.irqs[irq]++;
 -	/* block until irq becomes pending */
 -	xen_poll_irq(irq);
+ 	kstat_irqs_this_cpu(irq_to_desc(irq))++;
  
  out:
 -	unspinning_lock(xl);
 +	raw_local_irq_restore(flags);
 +	unspinning_lock(xl, prev);
 +	spin_time_accum_blocked(start);
 +
  	return ret;
  }
  

Re: linux-next: manual merge of the sparseirq tree

[Jeremy Fitzhardinge]

Stephen Rothwell wrote:
> [Ingo, thanks for the heads up about this ...]
>
> Hi Ingo,
>
> Today's linux-next merge of the sparseirq tree got a conflict in
> arch/x86/xen/spinlock.c between commit
> 168d2f464ab9860f0d1e66cf1f9684973222f1c6 ("xen: save previous spinlock
> when blocking") from the x86 tree and commit
> fb6dc57946f9ebfeac546dd0698d9f065c191668 ("x86: move kstat_irqs from
> kstat to irq_desc") from the sparseirq tree.
>
> I fixed it up (see below) and can carry it.
>   

This is a diff-diff?  The change:

- 	kstat_this_cpu.irqs[irq]++;
 -	/* block until irq becomes pending */
 -	xen_poll_irq(irq);
+ 	kstat_irqs_this_cpu(irq_to_desc(irq))++;


looks fine.

Re: linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 522 bytes --]

Hi Jeremy,

On Mon, 25 Aug 2008 12:37:27 -0700 Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>
> This is a diff-diff?  The change:

Its from "git diff" of the merge itself (which implies the -cc option of
git diff).

> 
> - 	kstat_this_cpu.irqs[irq]++;
>  -	/* block until irq becomes pending */
>  -	xen_poll_irq(irq);
> + 	kstat_irqs_this_cpu(irq_to_desc(irq))++;
> 
> 
> looks fine.

Thanks.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 551 bytes --]

Hi Ingo,

Today's linux-next merge of the sparseirq tree got a trivial conflict in
arch/x86/kvm/vmx.h and include/asm-x86/msr-index.h between commit
f5dd05a1980ab0770f1889fbc3ad2daa5ff89d6d ("x86: Move FEATURE_CONTROL bits
to msr-index.h") from the kvm tree and commit
315a6558f30a264c88274fa70626615d1c7851c7 ("x86: move VMX MSRs to
msr-index.h") from the sparseirq tree.

Just context changes.  I fixed it up and can carry the fix.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 4339 bytes --]

Hi Inog,

Today's linux-next merge of the sparseirq tree got a conflict in
arch/x86/kernel/ptrace.c between commit
393084e9e8c20c63cef574a6466ec0e59c711727 ("x86: ioperm user_regset") from
the x86 tree and commit 93fa7636dfdc059b25df148f230c0991096afdef ("x86,
ptrace: PEBS support") from the sparseirq tree.

Mostly overlapping additions/changes.  I fixed it up (see below) and can
carry the fix.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

diff --cc arch/x86/kernel/ptrace.c
index 6b7ab0b,58ce4b5..0000000
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@@ -557,41 -554,87 +557,110 @@@ static int ptrace_set_debugreg(struct t
  	return 0;
  }
  
 +/*
 + * These access the current or another (stopped) task's io permission
 + * bitmap for debugging or core dump.
 + */
 +static int ioperm_active(struct task_struct *target,
 +			 const struct user_regset *regset)
 +{
 +	return target->thread.io_bitmap_max / regset->size;
 +}
 +
 +static int ioperm_get(struct task_struct *target,
 +		      const struct user_regset *regset,
 +		      unsigned int pos, unsigned int count,
 +		      void *kbuf, void __user *ubuf)
 +{
 +	if (!target->thread.io_bitmap_ptr)
 +		return -ENXIO;
 +
 +	return user_regset_copyout(&pos, &count, &kbuf, &ubuf,
 +				   target->thread.io_bitmap_ptr,
 +				   0, IO_BITMAP_BYTES);
 +}
 +
- #ifdef X86_BTS
+ #ifdef CONFIG_X86_PTRACE_BTS
+ /*
+  * The configuration for a particular BTS hardware implementation.
+  */
+ struct bts_configuration {
+ 	/* the size of a BTS record in bytes; at most BTS_MAX_RECORD_SIZE */
+ 	unsigned char  sizeof_bts;
+ 	/* the size of a field in the BTS record in bytes */
+ 	unsigned char  sizeof_field;
+ 	/* a bitmask to enable/disable BTS in DEBUGCTL MSR */
+ 	unsigned long debugctl_mask;
+ };
+ static struct bts_configuration bts_cfg;
+ 
+ #define BTS_MAX_RECORD_SIZE (8 * 3)
+ 
+ 
+ /*
+  * Branch Trace Store (BTS) uses the following format. Different
+  * architectures vary in the size of those fields.
+  * - source linear address
+  * - destination linear address
+  * - flags
+  *
+  * Later architectures use 64bit pointers throughout, whereas earlier
+  * architectures use 32bit pointers in 32bit mode.
+  *
+  * We compute the base address for the first 8 fields based on:
+  * - the field size stored in the DS configuration
+  * - the relative field position
+  *
+  * In order to store additional information in the BTS buffer, we use
+  * a special source address to indicate that the record requires
+  * special interpretation.
+  *
+  * Netburst indicated via a bit in the flags field whether the branch
+  * was predicted; this is ignored.
+  */
  
- static int ptrace_bts_get_size(struct task_struct *child)
+ enum bts_field {
+ 	bts_from = 0,
+ 	bts_to,
+ 	bts_flags,
+ 
+ 	bts_escape = (unsigned long)-1,
+ 	bts_qual = bts_to,
+ 	bts_jiffies = bts_flags
+ };
+ 
+ static inline unsigned long bts_get(const char *base, enum bts_field field)
  {
- 	if (!child->thread.ds_area_msr)
- 		return -ENXIO;
+ 	base += (bts_cfg.sizeof_field * field);
+ 	return *(unsigned long *)base;
+ }
  
- 	return ds_get_bts_index((void *)child->thread.ds_area_msr);
+ static inline void bts_set(char *base, enum bts_field field, unsigned long val)
+ {
+ 	base += (bts_cfg.sizeof_field * field);;
+ 	(*(unsigned long *)base) = val;
  }
  
- static int ptrace_bts_read_record(struct task_struct *child,
- 				  long index,
+ /*
+  * Translate a BTS record from the raw format into the bts_struct format
+  *
+  * out (out): bts_struct interpretation
+  * raw: raw BTS record
+  */
+ static void ptrace_bts_translate_record(struct bts_struct *out, const void *raw)
+ {
+ 	memset(out, 0, sizeof(*out));
+ 	if (bts_get(raw, bts_from) == bts_escape) {
+ 		out->qualifier       = bts_get(raw, bts_qual);
+ 		out->variant.jiffies = bts_get(raw, bts_jiffies);
+ 	} else {
+ 		out->qualifier = BTS_BRANCH;
+ 		out->variant.lbr.from_ip = bts_get(raw, bts_from);
+ 		out->variant.lbr.to_ip   = bts_get(raw, bts_to);
+ 	}
+ }
+ 
+ static int ptrace_bts_read_record(struct task_struct *child, size_t index,
  				  struct bts_struct __user *out)
  {
  	struct bts_struct ret;

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: manual merge of the sparseirq tree

[Roland McGrath]

The ioperm patch (tip/x86/ptrace) adds the new ioperm_* functions, which
could go anywhere in the file (before the *_regsets[] decls).  It might be
simplest to just change the patch to put them earlier or later where the
patch text is not near the lines touched by BTS patches.


Thanks,
Roland

Re: linux-next: manual merge of the sparseirq tree

[Ingo Molnar]

* Roland McGrath <roland@redhat.com> wrote:

> The ioperm patch (tip/x86/ptrace) adds the new ioperm_* functions, 
> which could go anywhere in the file (before the *_regsets[] decls).  
> It might be simplest to just change the patch to put them earlier or 
> later where the patch text is not near the lines touched by BTS 
> patches.

yeah - but these are old commits, we better not upset them.

I've carried that conflict fixup for a long time and made it visible 
because i updated sparseirq tree to propagate a fix but the x86 tree for 
linux-next was at an older version (with PEBS topic not integrated yet) 
- so Stephen got to see this conflict.

It should disappear on the next integration run this week. (the 
frequency of integration runs is weekly - the frequency of bugfix/hotfix 
propagation is daily/hourly)

Thanks Stephen for taking care of it.

	Ingo

linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 407 bytes --]

Hi Ingo,

Today's linux-next merge of the sparseirq tree got a lot of conflicts
with various other trees.  For today, I will drop it and maybe have
another look tomorrow when things are (hopefully) less busy.  If you
could fix up any internal-to-tip and Linus conflicts, that would be great.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: manual merge of the sparseirq tree

[Ingo Molnar]

* Stephen Rothwell <sfr@canb.auug.org.au> wrote:

> Hi Ingo,
> 
> Today's linux-next merge of the sparseirq tree got a lot of conflicts 
> with various other trees.  For today, I will drop it and maybe have 
> another look tomorrow when things are (hopefully) less busy. [...]

okay, agreed - i've just reset auto-sparseirq-next to Linus and will 
update it with a merged up tree. (which we have in the works)

> [...] If you could fix up any internal-to-tip and Linus conflicts, 
> that would be great.

no other topic should be conflicting that i know of.

	Ingo

Re: linux-next: manual merge of the sparseirq tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 428 bytes --]

Hi Ingo,

On Thu, 16 Oct 2008 10:13:47 +0200 Ingo Molnar <mingo@elte.hu> wrote:
>
> okay, agreed - i've just reset auto-sparseirq-next to Linus and will 
> update it with a merged up tree. (which we have in the works)

OK, thanks.

> no other topic should be conflicting that i know of.

I was guessing :-)

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]