Re: [PATCH 3/4] string: introduce memchr_inv

Re: [PATCH 3/4] string: introduce memchr_inv

[Andrew Morton]

On Tue, 23 Aug 2011 01:29:07 +0900
Akinobu Mita <akinobu.mita@gmail.com> wrote:

> memchr_inv() is mainly used to check whether the whole buffer is filled
> with just a specified byte.
> 
> The function name and prototype are stolen from logfs and the
> implementation is from SLUB.
> 
> ...
>
> +/**
> + * memchr_inv - Find a character in an area of memory.
> + * @s: The memory area
> + * @c: The byte to search for
> + * @n: The size of the area.

This text seems to be stolen from memchr().  I guess it's close enough.

> + * returns the address of the first character other than @c, or %NULL
> + * if the whole buffer contains just @c.
> + */
> +void *memchr_inv(const void *start, int c, size_t bytes)
> +{
> +	u8 value = c;
> +	u64 value64;
> +	unsigned int words, prefix;
> +
> +	if (bytes <= 16)
> +		return check_bytes8(start, value, bytes);
> +
> +	value64 = value | value << 8 | value << 16 | value << 24;
> +	value64 = (value64 & 0xffffffff) | value64 << 32;
> +	prefix = 8 - ((unsigned long)start) % 8;
> +
> +	if (prefix) {
> +		u8 *r = check_bytes8(start, value, prefix);
> +		if (r)
> +			return r;
> +		start += prefix;
> +		bytes -= prefix;
> +	}
> +
> +	words = bytes / 8;
> +
> +	while (words) {
> +		if (*(u64 *)start != value64)

OK, problem.  This will explode if passed a misaligned address on
certain (non-x86) architectures.  This is nasty because people will
develop and test code on x86 and it works.  Much later, the
alpha/ia64/etc guys discover the problem.

One fix would be to use get_unaligned().  This might be slow on some
architectures, I don't know.  Another fix is to restrict the caller's
alignment freedom; document this and add a runtime WARN_ON().

> +			return check_bytes8(start, value, 8);
> +		start += 8;
> +		words--;
> +	}
> +
> +	return check_bytes8(start, value, bytes % 8);
> +}
> +EXPORT_SYMBOL(memchr_inv);

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/4] string: introduce memchr_inv

[Andrew Morton]

On Tue, 23 Aug 2011 01:29:07 +0900
Akinobu Mita <akinobu.mita@gmail.com> wrote:

> memchr_inv() is mainly used to check whether the whole buffer is filled
> with just a specified byte.
> 
> The function name and prototype are stolen from logfs and the
> implementation is from SLUB.
> 
> ...
>
> +/**
> + * memchr_inv - Find a character in an area of memory.
> + * @s: The memory area
> + * @c: The byte to search for
> + * @n: The size of the area.

This text seems to be stolen from memchr().  I guess it's close enough.

> + * returns the address of the first character other than @c, or %NULL
> + * if the whole buffer contains just @c.
> + */
> +void *memchr_inv(const void *start, int c, size_t bytes)
> +{
> +	u8 value = c;
> +	u64 value64;
> +	unsigned int words, prefix;
> +
> +	if (bytes <= 16)
> +		return check_bytes8(start, value, bytes);
> +
> +	value64 = value | value << 8 | value << 16 | value << 24;
> +	value64 = (value64 & 0xffffffff) | value64 << 32;
> +	prefix = 8 - ((unsigned long)start) % 8;
> +
> +	if (prefix) {
> +		u8 *r = check_bytes8(start, value, prefix);
> +		if (r)
> +			return r;
> +		start += prefix;
> +		bytes -= prefix;
> +	}
> +
> +	words = bytes / 8;
> +
> +	while (words) {
> +		if (*(u64 *)start != value64)

OK, problem.  This will explode if passed a misaligned address on
certain (non-x86) architectures.  This is nasty because people will
develop and test code on x86 and it works.  Much later, the
alpha/ia64/etc guys discover the problem.

One fix would be to use get_unaligned().  This might be slow on some
architectures, I don't know.  Another fix is to restrict the caller's
alignment freedom; document this and add a runtime WARN_ON().

> +			return check_bytes8(start, value, 8);
> +		start += 8;
> +		words--;
> +	}
> +
> +	return check_bytes8(start, value, bytes % 8);
> +}
> +EXPORT_SYMBOL(memchr_inv);

Re: [PATCH 3/4] string: introduce memchr_inv

[Andrew Morton]

On Mon, 22 Aug 2011 13:52:18 -0700
Andrew Morton <akpm@linux-foundation.org> wrote:

> > +	value64 = value | value << 8 | value << 16 | value << 24;
> > +	value64 = (value64 & 0xffffffff) | value64 << 32;
> > +	prefix = 8 - ((unsigned long)start) % 8;
> > +
> > +	if (prefix) {
> > +		u8 *r = check_bytes8(start, value, prefix);
> > +		if (r)
> > +			return r;
> > +		start += prefix;
> > +		bytes -= prefix;
> > +	}
> > +
> > +	words = bytes / 8;
> > +
> > +	while (words) {
> > +		if (*(u64 *)start != value64)
> 
> OK, problem.  This will explode if passed a misaligned address on
> certain (non-x86) architectures.

pls ignore.  As Marcin points out, I can't read.

Re: [PATCH 3/4] string: introduce memchr_inv

[Andrew Morton]

On Mon, 22 Aug 2011 13:52:18 -0700
Andrew Morton <akpm@linux-foundation.org> wrote:

> > +	value64 = value | value << 8 | value << 16 | value << 24;
> > +	value64 = (value64 & 0xffffffff) | value64 << 32;
> > +	prefix = 8 - ((unsigned long)start) % 8;
> > +
> > +	if (prefix) {
> > +		u8 *r = check_bytes8(start, value, prefix);
> > +		if (r)
> > +			return r;
> > +		start += prefix;
> > +		bytes -= prefix;
> > +	}
> > +
> > +	words = bytes / 8;
> > +
> > +	while (words) {
> > +		if (*(u64 *)start != value64)
> 
> OK, problem.  This will explode if passed a misaligned address on
> certain (non-x86) architectures.

pls ignore.  As Marcin points out, I can't read.

Re: [PATCH 3/4] string: introduce memchr_inv

[Eric Dumazet]

Le lundi 22 août 2011 à 13:52 -0700, Andrew Morton a écrit :
> On Tue, 23 Aug 2011 01:29:07 +0900
> Akinobu Mita <akinobu.mita@gmail.com> wrote:
> 
> > memchr_inv() is mainly used to check whether the whole buffer is filled
> > with just a specified byte.
> > 
> > The function name and prototype are stolen from logfs and the
> > implementation is from SLUB.
> > 
> > ...
> >
> > +/**
> > + * memchr_inv - Find a character in an area of memory.
> > + * @s: The memory area
> > + * @c: The byte to search for
> > + * @n: The size of the area.
> 
> This text seems to be stolen from memchr().  I guess it's close enough.
> 
> > + * returns the address of the first character other than @c, or %NULL
> > + * if the whole buffer contains just @c.
> > + */
> > +void *memchr_inv(const void *start, int c, size_t bytes)
> > +{
> > +	u8 value = c;
> > +	u64 value64;
> > +	unsigned int words, prefix;
> > +
> > +	if (bytes <= 16)
> > +		return check_bytes8(start, value, bytes);
> > +
> > +	value64 = value | value << 8 | value << 16 | value << 24;
> > +	value64 = (value64 & 0xffffffff) | value64 << 32;
> > +	prefix = 8 - ((unsigned long)start) % 8;
> > +

<snip>

> > +	if (prefix) {
> > +		u8 *r = check_bytes8(start, value, prefix);
> > +		if (r)
> > +			return r;
> > +		start += prefix;
> > +		bytes -= prefix;
> > +	}

</snip>

Please note Andrew the previous code just make sure 'start' is aligned
on 8 bytes boundary. (It is suboptimal because if 'start' was already
aligned, we call the slow check_bytes(start, value, 8))

Code should probably do

prefix = (unsigned long)start % 8;
if (prefix) {
	prefix = 8 - prefix;
	r = check_bytes8(start, value, prefix);
	...



> > +
> > +	words = bytes / 8;
> > +
> > +	while (words) {
> > +		if (*(u64 *)start != value64)



--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/4] string: introduce memchr_inv

[Eric Dumazet]

Le lundi 22 août 2011 à 13:52 -0700, Andrew Morton a écrit :
> On Tue, 23 Aug 2011 01:29:07 +0900
> Akinobu Mita <akinobu.mita@gmail.com> wrote:
> 
> > memchr_inv() is mainly used to check whether the whole buffer is filled
> > with just a specified byte.
> > 
> > The function name and prototype are stolen from logfs and the
> > implementation is from SLUB.
> > 
> > ...
> >
> > +/**
> > + * memchr_inv - Find a character in an area of memory.
> > + * @s: The memory area
> > + * @c: The byte to search for
> > + * @n: The size of the area.
> 
> This text seems to be stolen from memchr().  I guess it's close enough.
> 
> > + * returns the address of the first character other than @c, or %NULL
> > + * if the whole buffer contains just @c.
> > + */
> > +void *memchr_inv(const void *start, int c, size_t bytes)
> > +{
> > +	u8 value = c;
> > +	u64 value64;
> > +	unsigned int words, prefix;
> > +
> > +	if (bytes <= 16)
> > +		return check_bytes8(start, value, bytes);
> > +
> > +	value64 = value | value << 8 | value << 16 | value << 24;
> > +	value64 = (value64 & 0xffffffff) | value64 << 32;
> > +	prefix = 8 - ((unsigned long)start) % 8;
> > +

<snip>

> > +	if (prefix) {
> > +		u8 *r = check_bytes8(start, value, prefix);
> > +		if (r)
> > +			return r;
> > +		start += prefix;
> > +		bytes -= prefix;
> > +	}

</snip>

Please note Andrew the previous code just make sure 'start' is aligned
on 8 bytes boundary. (It is suboptimal because if 'start' was already
aligned, we call the slow check_bytes(start, value, 8))

Code should probably do

prefix = (unsigned long)start % 8;
if (prefix) {
	prefix = 8 - prefix;
	r = check_bytes8(start, value, prefix);
	...



> > +
> > +	words = bytes / 8;
> > +
> > +	while (words) {
> > +		if (*(u64 *)start != value64)