Hi!
This set moves to a global header file macros which I find
very useful and worth popularising. The basic problem is
that since C bitfields are not very dependable accessing
subfields of registers becomes slightly inconvenient.
It is nice to have the necessary mask and shift operations
wrapped in a macro. It is also nice to have that macro
compute the shift amount based on the mask automatically.
My implementation follows what Felix Fietkau has done in
mt76. Hannes Frederic Sowa suggested more use of standard
Linux/GCC functions. Since the RFC I've also added a
compile-time check to validate that the value passed to
setters fits in the mask.
I attempted the use of static inlines instead of macros
but it makes GCC < 6.0 barf at the BUILD_BUG_ON()s.
I also noticed that forcing arguments to be u32 for inlines
makes the compiler use 32bit arithmetic where it could
get away with 64bit before (on 64bit machines, obviously).
That's a potential performance concern but probably not
a very practical one today. Apart from looking "cleaner"
static inlines would have the advantage that we could #undef
the auxiliary macros at the end of the header.
v3:
Build bot caught a build failure with -Os set. AFAICT gcc
did not handle temporary variable I put in the macro
expression too well. I work around that by defining
__BUILD_BUG_ON_NOT_POWER_OF_2 and using it instead of
BUILD_BUG_ON(!tmp || is_power_of_2(tmp)).
Please review and advise on improvements.
If accepted I think would be best to push this through
Kalle's tree, since the only existing user is in
drivers/net/wireless/.
v3:
- don't use variables in statement expressions;
- use __BUILD_BUG_ON_NOT_POWER_OF_2.
v2:
- change Felix's email address.
Jakub Kicinski (2):
add basic register-field manipulation macros
mt7601u: use linux/bitfield.h
drivers/net/wireless/mediatek/mt7601u/dma.h | 2 -
drivers/net/wireless/mediatek/mt7601u/mt7601u.h | 5 +-
drivers/net/wireless/mediatek/mt7601u/util.h | 77 -------------------------
include/linux/bitfield.h | 58 +++++++++++++++++++
include/linux/bug.h | 3 +
5 files changed, 65 insertions(+), 80 deletions(-)
delete mode 100644 drivers/net/wireless/mediatek/mt7601u/util.h
create mode 100644 include/linux/bitfield.h
--
1.9.1
C bitfields are problematic and best avoided. Developers
interacting with hardware registers find themselves searching
for easy-to-use alternatives. Common approach is to define
structures or sets of macros containing mask and shift pair.
Operations on the register are then performed as follows:
field = (reg >> shift) & mask;
reg &= ~(mask << shift);
reg |= (field & mask) << shift;
Defining shift and mask separately is tedious. Ivo van Doorn
came up with an idea of computing them at compilation time
based on a single shifted mask (later refined by Felix) which
can be used like this:
#define REG_FIELD 0x000ff000
field = FIELD_GET(REG_FIELD, reg);
reg &= ~REG_FIELD;
reg |= FIELD_PUT(REG_FIELD, field);
FIELD_{GET,PUT} macros take care of finding out what the
appropriate shift is based on compilation time ffs operation.
GENMASK can be used to define registers (which is usually
less error-prone and easier to match with datasheets).
This approach is the most convenient I've seen so to limit code
multiplication let's move the macros to a global header file.
Signed-off-by: Jakub Kicinski <[email protected]>
---
include/linux/bitfield.h | 58 ++++++++++++++++++++++++++++++++++++++++++++++++
include/linux/bug.h | 3 +++
2 files changed, 61 insertions(+)
create mode 100644 include/linux/bitfield.h
diff --git a/include/linux/bitfield.h b/include/linux/bitfield.h
new file mode 100644
index 000000000000..d6a36c3c1775
--- /dev/null
+++ b/include/linux/bitfield.h
@@ -0,0 +1,58 @@
+/*
+ * Copyright (C) 2014 Felix Fietkau <[email protected]>
+ * Copyright (C) 2004 - 2009 Ivo van Doorn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2
+ * as published by the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _LINUX_BITFIELD_H
+#define _LINUX_BITFIELD_H
+
+#include <asm/types.h>
+#include <linux/bug.h>
+#include <linux/log2.h>
+
+#define _bf_shf(x) (__builtin_ffsll(x) - 1)
+
+#define _BF_FIELD_CHECK(_mask, _val) \
+ ({ \
+ BUILD_BUG_ON(!(_mask)); \
+ BUILD_BUG_ON(__builtin_constant_p(_val) ? \
+ ~((_mask) >> _bf_shf(_mask)) & (_val) : \
+ 0); \
+ __BUILD_BUG_ON_NOT_POWER_OF_2((_mask) + \
+ (1ULL << _bf_shf(_mask))); \
+ })
+
+#define FIELD_PUT(_mask, _val) \
+ ({ \
+ _BF_FIELD_CHECK(_mask, _val); \
+ ((u32)(_val) << _bf_shf(_mask)) & (_mask); \
+ })
+
+#define FIELD_GET(_mask, _val) \
+ ({ \
+ _BF_FIELD_CHECK(_mask, 0); \
+ (u32)(((_val) & (_mask)) >> _bf_shf(_mask)); \
+ })
+
+#define FIELD_PUT64(_mask, _val) \
+ ({ \
+ _BF_FIELD_CHECK(_mask, _val); \
+ ((u64)(_val) << _bf_shf(_mask)) & (_mask); \
+ })
+
+#define FIELD_GET64(_mask, _val) \
+ ({ \
+ _BF_FIELD_CHECK(_mask, 0); \
+ (u64)(((_val) & (_mask)) >> _bf_shf(_mask)); \
+ })
+
+#endif
diff --git a/include/linux/bug.h b/include/linux/bug.h
index e51b0709e78d..bba5bdae1681 100644
--- a/include/linux/bug.h
+++ b/include/linux/bug.h
@@ -13,6 +13,7 @@ enum bug_trap_type {
struct pt_regs;
#ifdef __CHECKER__
+#define __BUILD_BUG_ON_NOT_POWER_OF_2(n) (0)
#define BUILD_BUG_ON_NOT_POWER_OF_2(n) (0)
#define BUILD_BUG_ON_ZERO(e) (0)
#define BUILD_BUG_ON_NULL(e) ((void*)0)
@@ -24,6 +25,8 @@ struct pt_regs;
#else /* __CHECKER__ */
/* Force a compilation error if a constant expression is not a power of 2 */
+#define __BUILD_BUG_ON_NOT_POWER_OF_2(n) \
+ BUILD_BUG_ON(((n) & ((n) - 1)) != 0)
#define BUILD_BUG_ON_NOT_POWER_OF_2(n) \
BUILD_BUG_ON((n) == 0 || (((n) & ((n) - 1)) != 0))
--
1.9.1
On Tue, 5 Jul 2016 10:56:13 +0000, David Laight wrote:
> From: Jakub Kicinski
> > Sent: 01 July 2016 22:27
> >
> > C bitfields are problematic and best avoided. Developers
> > interacting with hardware registers find themselves searching
> > for easy-to-use alternatives. Common approach is to define
> > structures or sets of macros containing mask and shift pair.
> > Operations on the register are then performed as follows:
> >
> > field = (reg >> shift) & mask;
> >
> > reg &= ~(mask << shift);
> > reg |= (field & mask) << shift;
> >
> > Defining shift and mask separately is tedious. Ivo van Doorn
> > came up with an idea of computing them at compilation time
> > based on a single shifted mask (later refined by Felix) which
> > can be used like this:
> >
> > #define REG_FIELD 0x000ff000
> >
> > field = FIELD_GET(REG_FIELD, reg);
> >
> > reg &= ~REG_FIELD;
> > reg |= FIELD_PUT(REG_FIELD, field);
>
> My problem with these sort of 'helpers' is that they make it much harder
> to read code unless you happen to know exactly what the helpers do.
> Unexpected issues (like values being sign extended) can be hard to spot.
>
> A lot of the time you can make things simpler by not doing the shifts
> (ie define shifted constants).
I think creating a standard set of macros in a global header file is
actually helping the problems you raise. One is much more likely to
know exactly what a common macro is doing than some driver-specific
ad hoc macro. Common macros are also much more likely to be well
tested making "unexpected issues" less likely to appear.
Defining constants is fine in 20% of cases when you have only a small
set of meaningful values (e.g. what to do for a 8 bit delay or
priority field?) Besides when fields are not aligned to 4 bits it's
hard to tell from the shifted value what the base was.
Use the newly added linux/bitfield.h.
Signed-off-by: Jakub Kicinski <[email protected]>
---
drivers/net/wireless/mediatek/mt7601u/dma.h | 2 -
drivers/net/wireless/mediatek/mt7601u/mt7601u.h | 5 +-
drivers/net/wireless/mediatek/mt7601u/util.h | 77 -------------------------
3 files changed, 4 insertions(+), 80 deletions(-)
delete mode 100644 drivers/net/wireless/mediatek/mt7601u/util.h
diff --git a/drivers/net/wireless/mediatek/mt7601u/dma.h b/drivers/net/wireless/mediatek/mt7601u/dma.h
index 978e8a90b87f..166ac71905d2 100644
--- a/drivers/net/wireless/mediatek/mt7601u/dma.h
+++ b/drivers/net/wireless/mediatek/mt7601u/dma.h
@@ -18,8 +18,6 @@
#include <asm/unaligned.h>
#include <linux/skbuff.h>
-#include "util.h"
-
#define MT_DMA_HDR_LEN 4
#define MT_RX_INFO_LEN 4
#define MT_FCE_INFO_LEN 4
diff --git a/drivers/net/wireless/mediatek/mt7601u/mt7601u.h b/drivers/net/wireless/mediatek/mt7601u/mt7601u.h
index 428bd2f10b7b..5ef62e02ce66 100644
--- a/drivers/net/wireless/mediatek/mt7601u/mt7601u.h
+++ b/drivers/net/wireless/mediatek/mt7601u/mt7601u.h
@@ -15,6 +15,7 @@
#ifndef MT7601U_H
#define MT7601U_H
+#include <linux/bitfield.h>
#include <linux/kernel.h>
#include <linux/device.h>
#include <linux/mutex.h>
@@ -24,7 +25,6 @@
#include <linux/debugfs.h>
#include "regs.h"
-#include "util.h"
#define MT_CALIBRATE_INTERVAL (4 * HZ)
@@ -282,6 +282,9 @@ struct mt7601u_rxwi;
extern const struct ieee80211_ops mt7601u_ops;
+#define MT76_SET FIELD_PUT
+#define MT76_GET FIELD_GET
+
void mt7601u_init_debugfs(struct mt7601u_dev *dev);
u32 mt7601u_rr(struct mt7601u_dev *dev, u32 offset);
diff --git a/drivers/net/wireless/mediatek/mt7601u/util.h b/drivers/net/wireless/mediatek/mt7601u/util.h
deleted file mode 100644
index b89140bf1210..000000000000
--- a/drivers/net/wireless/mediatek/mt7601u/util.h
+++ /dev/null
@@ -1,77 +0,0 @@
-/*
- * Copyright (C) 2014 Felix Fietkau <[email protected]>
- * Copyright (C) 2004 - 2009 Ivo van Doorn <[email protected]>
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2
- * as published by the Free Software Foundation
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- */
-
-#ifndef __MT76_UTIL_H
-#define __MT76_UTIL_H
-
-/*
- * Power of two check, this will check
- * if the mask that has been given contains and contiguous set of bits.
- * Note that we cannot use the is_power_of_2() function since this
- * check must be done at compile-time.
- */
-#define is_power_of_two(x) ( !((x) & ((x)-1)) )
-#define low_bit_mask(x) ( ((x)-1) & ~(x) )
-#define is_valid_mask(x) is_power_of_two(1LU + (x) + low_bit_mask(x))
-
-/*
- * Macros to find first set bit in a variable.
- * These macros behave the same as the __ffs() functions but
- * the most important difference that this is done during
- * compile-time rather then run-time.
- */
-#define compile_ffs2(__x) \
- __builtin_choose_expr(((__x) & 0x1), 0, 1)
-
-#define compile_ffs4(__x) \
- __builtin_choose_expr(((__x) & 0x3), \
- (compile_ffs2((__x))), \
- (compile_ffs2((__x) >> 2) + 2))
-
-#define compile_ffs8(__x) \
- __builtin_choose_expr(((__x) & 0xf), \
- (compile_ffs4((__x))), \
- (compile_ffs4((__x) >> 4) + 4))
-
-#define compile_ffs16(__x) \
- __builtin_choose_expr(((__x) & 0xff), \
- (compile_ffs8((__x))), \
- (compile_ffs8((__x) >> 8) + 8))
-
-#define compile_ffs32(__x) \
- __builtin_choose_expr(((__x) & 0xffff), \
- (compile_ffs16((__x))), \
- (compile_ffs16((__x) >> 16) + 16))
-
-/*
- * This macro will check the requirements for the FIELD{8,16,32} macros
- * The mask should be a constant non-zero contiguous set of bits which
- * does not exceed the given typelimit.
- */
-#define FIELD_CHECK(__mask) \
- BUILD_BUG_ON(!(__mask) || !is_valid_mask(__mask))
-
-#define MT76_SET(_mask, _val) \
- ({ \
- FIELD_CHECK(_mask); \
- (((u32) (_val)) << compile_ffs32(_mask)) & _mask; \
- })
-
-#define MT76_GET(_mask, _val) \
- ({ \
- FIELD_CHECK(_mask); \
- (u32) (((_val) & _mask) >> compile_ffs32(_mask)); \
- })
-
-#endif
--
1.9.1
From: Jakub Kicinski
> Sent: 01 July 2016 22:27
>
> C bitfields are problematic and best avoided. Developers
> interacting with hardware registers find themselves searching
> for easy-to-use alternatives. Common approach is to define
> structures or sets of macros containing mask and shift pair.
> Operations on the register are then performed as follows:
>
> field = (reg >> shift) & mask;
>
> reg &= ~(mask << shift);
> reg |= (field & mask) << shift;
>
> Defining shift and mask separately is tedious. Ivo van Doorn
> came up with an idea of computing them at compilation time
> based on a single shifted mask (later refined by Felix) which
> can be used like this:
>
> #define REG_FIELD 0x000ff000
>
> field = FIELD_GET(REG_FIELD, reg);
>
> reg &= ~REG_FIELD;
> reg |= FIELD_PUT(REG_FIELD, field);
My problem with these sort of 'helpers' is that they make it much harder
to read code unless you happen to know exactly what the helpers do.
Unexpected issues (like values being sign extended) can be hard to spot.
A lot of the time you can make things simpler by not doing the shifts
(ie define shifted constants).
David
On 01.07.2016 23:26, Jakub Kicinski wrote:
> C bitfields are problematic and best avoided. Developers
> interacting with hardware registers find themselves searching
> for easy-to-use alternatives. Common approach is to define
> structures or sets of macros containing mask and shift pair.
> Operations on the register are then performed as follows:
>
> field = (reg >> shift) & mask;
>
> reg &= ~(mask << shift);
> reg |= (field & mask) << shift;
>
> Defining shift and mask separately is tedious. Ivo van Doorn
> came up with an idea of computing them at compilation time
> based on a single shifted mask (later refined by Felix) which
> can be used like this:
>
> #define REG_FIELD 0x000ff000
>
> field = FIELD_GET(REG_FIELD, reg);
>
> reg &= ~REG_FIELD;
> reg |= FIELD_PUT(REG_FIELD, field);
>
> FIELD_{GET,PUT} macros take care of finding out what the
> appropriate shift is based on compilation time ffs operation.
>
> GENMASK can be used to define registers (which is usually
> less error-prone and easier to match with datasheets).
>
> This approach is the most convenient I've seen so to limit code
> multiplication let's move the macros to a global header file.
>
> Signed-off-by: Jakub Kicinski <[email protected]>
> ---
> include/linux/bitfield.h | 58 ++++++++++++++++++++++++++++++++++++++++++++++++
> include/linux/bug.h | 3 +++
> 2 files changed, 61 insertions(+)
> create mode 100644 include/linux/bitfield.h
>
> diff --git a/include/linux/bitfield.h b/include/linux/bitfield.h
> new file mode 100644
> index 000000000000..d6a36c3c1775
> --- /dev/null
> +++ b/include/linux/bitfield.h
> @@ -0,0 +1,58 @@
> +/*
> + * Copyright (C) 2014 Felix Fietkau <[email protected]>
> + * Copyright (C) 2004 - 2009 Ivo van Doorn <[email protected]>
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2
> + * as published by the Free Software Foundation
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
> +
> +#ifndef _LINUX_BITFIELD_H
> +#define _LINUX_BITFIELD_H
> +
> +#include <asm/types.h>
> +#include <linux/bug.h>
> +#include <linux/log2.h>
> +
> +#define _bf_shf(x) (__builtin_ffsll(x) - 1)
> +
> +#define _BF_FIELD_CHECK(_mask, _val) \
> + ({ \
> + BUILD_BUG_ON(!(_mask)); \
> + BUILD_BUG_ON(__builtin_constant_p(_val) ? \
> + ~((_mask) >> _bf_shf(_mask)) & (_val) : \
> + 0); \
> + __BUILD_BUG_ON_NOT_POWER_OF_2((_mask) + \
> + (1ULL << _bf_shf(_mask))); \
> + })
> +
> +#define FIELD_PUT(_mask, _val) \
> + ({ \
> + _BF_FIELD_CHECK(_mask, _val); \
> + ((u32)(_val) << _bf_shf(_mask)) & (_mask); \
> + })
> +
> +#define FIELD_GET(_mask, _val) \
> + ({ \
> + _BF_FIELD_CHECK(_mask, 0); \
> + (u32)(((_val) & (_mask)) >> _bf_shf(_mask)); \
> + })
> +
> +#define FIELD_PUT64(_mask, _val) \
> + ({ \
> + _BF_FIELD_CHECK(_mask, _val); \
> + ((u64)(_val) << _bf_shf(_mask)) & (_mask); \
> + })
> +
> +#define FIELD_GET64(_mask, _val) \
> + ({ \
> + _BF_FIELD_CHECK(_mask, 0); \
> + (u64)(((_val) & (_mask)) >> _bf_shf(_mask)); \
> + })
> +
> +#endif
Cool! I think this is totally fine. Albeit one point: can you put your
commit comment into the header itself. People check header files more
often than git commits for documentation and it should be quickly
graspable by a developer if someone looks them up.
Thanks,
Hannes