From: Sascha Hauer <s.hauer@pengutronix.de>
To: Antony Pavlov <antonynpavlov@gmail.com>
Cc: barebox@lists.infradead.org
Subject: Re: [PATCH 1/4] byteorder: fix {BIG,LITTLE}_ENDIAN defines
Date: Thu, 28 Jun 2012 12:19:17 +0200 [thread overview]
Message-ID: <20120628101917.GU1623@pengutronix.de> (raw)
In-Reply-To: <1340876689-26303-2-git-send-email-antonynpavlov@gmail.com>
Hi Antony,
On Thu, Jun 28, 2012 at 01:44:46PM +0400, Antony Pavlov wrote:
> In the Linux kernel sources the only one byteorder macro
> can be defined (__BIG_ENDIAN or __LITTLE_ENDIAN) at a time.
>
> In barebox we have the __BIG_ENDIAN and __LITTLE_ENDIAN macros
> defined simultaneously introduced in
>
> commit 9ad1fe64abb12baac918ec177d9a52bbf2980d16
> Author: Baruch Siach <baruch@tkos.co.il>
> Date: Sun Jun 27 08:46:05 2010 +0300
>
> byteorder: add missing {BIG,LITTLE}_ENDIAN defines
>
> This fixes build warnings when testing __BYTE_ORDER
> of the other kin
>
> But in arch/mips/lib/libgcc.h (from Linux) we have
>
> #ifdef __BIG_ENDIAN
> struct DWstruct {
> int high, low;
> };
> #elif defined(__LITTLE_ENDIAN)
> struct DWstruct {
> int low, high;
> };
> #else
> #error I feel sick.
> #endif
>
> This means that regardless of current byteorder the big-endian
> DWstruct will be selected. By turn this breaks the __lshrdi3()
> function and the clocksource code on little-endian MIPS.
Additionally I suggest to add the following to include/common.h:
/*
* sanity check. The Linux Kernel defines only one of __LITTLE_ENDIAN and
* __BIG_ENDIAN. Endianess can then be tested with #ifdef __xx_ENDIAN. Userspace
* always defined both __LITTLE_ENDIAN and __BIG_ENDIAN and byteorder can then
* be tested with #if __BYTE_ORDER == __xx_ENDIAN.
*
* As we tend to use a lot of Kernel code in barebox we use the kernel way of
* determing the byte order. Make sure here that architecture code properly
* defines it.
*/
#include <asm/byteorder.h>
#if defined __LITTLE_ENDIAN && defined __BIG_ENDIAN
#error "both __LITTLE_ENDIAN and __BIG_ENDIAN are defined"
#endif
#if !defined __LITTLE_ENDIAN && !defined __BIG_ENDIAN
#error "None of __LITTLE_ENDIAN and __BIG_ENDIAN are defined"
#endif
> #define CRAMFS_24(x) ((swab32(x)) >> 8)
> diff --git a/include/envfs.h b/include/envfs.h
> index 67b8902..c053f19 100644
> --- a/include/envfs.h
> +++ b/include/envfs.h
> @@ -38,7 +38,7 @@ struct envfs_super {
> #error "No byte order defined in __BYTE_ORDER"
> #endif
>
> -#if __BYTE_ORDER == __LITTLE_ENDIAN
> +#ifdef __LITTLE_ENDIAN
Be careful. This is also included when building the barebox env tool.
Getting the endianess right is tricky. As mentioned above in Userspace
we have:
#define __LITTLE_ENDIAN 1234
#define __BIG_ENDIAN 4321
#define __BYTE_ORDER [__LITTLE_ENDIAN|__BIG_ENDIAN]
Whereas the kernel only defines either __LITTLE_ENDIAN or __BIG_ENDIAN
This means that #ifdef __LITTLE_ENDIAN in userspace will always be true
regardless of the endianess.
> diff --git a/scripts/compiler.h b/scripts/compiler.h
> index 53f84b6..dac7d46 100644
> --- a/scripts/compiler.h
> +++ b/scripts/compiler.h
> @@ -79,7 +79,7 @@ typedef uint32_t __u32;
> # define uswap_64(x) _uswap_64(x, )
> #endif
>
> -#if __BYTE_ORDER == __LITTLE_ENDIAN
> +#ifdef __LITTLE_ENDIAN
So this is also wrong.
> # define cpu_to_le16(x) (x)
> # define cpu_to_le32(x) (x)
> # define cpu_to_le64(x) (x)
> diff --git a/scripts/setupmbr/arch.h b/scripts/setupmbr/arch.h
> index a720dfe..1dc3f50 100644
> --- a/scripts/setupmbr/arch.h
> +++ b/scripts/setupmbr/arch.h
> @@ -32,7 +32,7 @@
> | (_x >> 56)); \
> })
>
> -#if __BYTE_ORDER == __BIG_ENDIAN
> +#ifdef __BIG_ENDIAN
ditto.
I suggest using the following patch which should do the things I
mentioned right.
Sascha
8<----------------------------------------------------------
Change byte order detection mechanism to kernel style
The Linux Kernel defines only one of __LITTLE_ENDIAN and
__BIG_ENDIAN. Endianess can then be tested with #ifdef __xx_ENDIAN. Userspace
always defined both __LITTLE_ENDIAN and __BIG_ENDIAN and byteorder can then
be tested with #if __BYTE_ORDER == __xx_ENDIAN.
As we tend to use a lot of Kernel code in barebox we switch to use the kernel
way of determing the byte order.
As this always causes a lot of confusion add a check to include/common.h to
make sure only one of __LITTLE_ENDIAN and __BIG_ENDIAN is defined.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
drivers/ata/disk_ata_drive.c | 2 +-
drivers/nor/cfi_flash.c | 10 ++++++----
include/common.h | 18 ++++++++++++++++++
include/cramfs/cramfs_fs.h | 8 ++------
include/envfs.h | 22 ++++++++++++++++++----
include/linux/byteorder/generic.h | 7 -------
include/usb/usb.h | 6 ++++--
7 files changed, 49 insertions(+), 24 deletions(-)
diff --git a/drivers/ata/disk_ata_drive.c b/drivers/ata/disk_ata_drive.c
index 4602af3..d5c5837 100644
--- a/drivers/ata/disk_ata_drive.c
+++ b/drivers/ata/disk_ata_drive.c
@@ -231,7 +231,7 @@ static void __maybe_unused ata_dump_id(uint16_t *id)
*/
static void ata_fix_endianess(uint16_t *buf, unsigned wds)
{
-#if __BYTE_ORDER == __BIG_ENDIAN
+#ifdef __BIG_ENDIAN
unsigned u;
for (u = 0; u < wds; u++)
diff --git a/drivers/nor/cfi_flash.c b/drivers/nor/cfi_flash.c
index 654e647..82c1cae 100644
--- a/drivers/nor/cfi_flash.c
+++ b/drivers/nor/cfi_flash.c
@@ -82,9 +82,9 @@ static void flash_add_byte (struct flash_info *info, cfiword_t * cword, uchar c)
return;
}
-#if __BYTE_ORDER == __BIG_ENDIAN
+#ifdef __BIG_ENDIAN
*cword = (*cword << 8) | c;
-#else
+#elif defined __LITTLE_ENDIAN
if (bankwidth_is_2(info))
*cword = (*cword >> 8) | (u16)c << 8;
@@ -92,6 +92,8 @@ static void flash_add_byte (struct flash_info *info, cfiword_t * cword, uchar c)
*cword = (*cword >> 8) | (u32)c << 24;
else if (bankwidth_is_8(info))
*cword = (*cword >> 8) | (u64)c << 56;
+#else
+#error "could not determine byte order"
#endif
}
@@ -167,7 +169,7 @@ static void flash_printqry (struct cfi_qry *qry)
uchar flash_read_uchar (struct flash_info *info, uint offset)
{
uchar *cp = flash_make_addr(info, 0, offset);
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+#if defined __LITTLE_ENDIAN
return flash_read8(cp);
#else
return flash_read8(cp + info->portwidth - 1);
@@ -195,7 +197,7 @@ static ulong flash_read_long (struct flash_info *info, flash_sect_t sect, uint o
debug ("addr[%x] = 0x%x\n", x, flash_read8(addr + x));
}
#endif
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+#if defined __LITTLE_ENDIAN
retval = ((flash_read8(addr) << 16) |
(flash_read8(addr + info->portwidth) << 24) |
(flash_read8(addr + 2 * info->portwidth)) |
diff --git a/include/common.h b/include/common.h
index d2347f8..9287a7d 100644
--- a/include/common.h
+++ b/include/common.h
@@ -34,6 +34,24 @@
#include <linux/stddef.h>
#include <asm/common.h>
+/*
+ * sanity check. The Linux Kernel defines only one of __LITTLE_ENDIAN and
+ * __BIG_ENDIAN. Endianess can then be tested with #ifdef __xx_ENDIAN. Userspace
+ * always defined both __LITTLE_ENDIAN and __BIG_ENDIAN and byteorder can then
+ * be tested with #if __BYTE_ORDER == __xx_ENDIAN.
+ *
+ * As we tend to use a lot of Kernel code in barebox we use the kernel way of
+ * determing the byte order. Make sure here that architecture code properly
+ * defines it.
+ */
+#include <asm/byteorder.h>
+#if defined __LITTLE_ENDIAN && defined __BIG_ENDIAN
+#error "both __LITTLE_ENDIAN and __BIG_ENDIAN are defined"
+#endif
+#if !defined __LITTLE_ENDIAN && !defined __BIG_ENDIAN
+#error "None of __LITTLE_ENDIAN and __BIG_ENDIAN are defined"
+#endif
+
#define pr_info(fmt, arg...) printf(fmt, ##arg)
#define pr_notice(fmt, arg...) printf(fmt, ##arg)
#define pr_err(fmt, arg...) printf(fmt, ##arg)
diff --git a/include/cramfs/cramfs_fs.h b/include/cramfs/cramfs_fs.h
index af2940b..8c53fc7 100644
--- a/include/cramfs/cramfs_fs.h
+++ b/include/cramfs/cramfs_fs.h
@@ -84,11 +84,7 @@ struct cramfs_super {
| CRAMFS_FLAG_WRONG_SIGNATURE \
| CRAMFS_FLAG_SHIFTED_ROOT_OFFSET )
-#ifndef __BYTE_ORDER
-#error "No byte order defined in __BYTE_ORDER"
-#endif
-
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+#ifdef __LITTLE_ENDIAN
#define CRAMFS_16(x) (x)
#define CRAMFS_24(x) (x)
#define CRAMFS_32(x) (x)
@@ -96,7 +92,7 @@ struct cramfs_super {
#define CRAMFS_GET_OFFSET(x) ((x)->offset)
#define CRAMFS_SET_OFFSET(x,y) ((x)->offset = (y))
#define CRAMFS_SET_NAMELEN(x,y) ((x)->namelen = (y))
-#elif __BYTE_ORDER ==__BIG_ENDIAN
+#elif defined __BIG_ENDIAN
#ifdef __KERNEL__
#define CRAMFS_16(x) swab16(x)
#define CRAMFS_24(x) ((swab32(x)) >> 8)
diff --git a/include/envfs.h b/include/envfs.h
index 67b8902..ba976d6 100644
--- a/include/envfs.h
+++ b/include/envfs.h
@@ -34,11 +34,25 @@ struct envfs_super {
uint32_t sb_crc; /* crc for the superblock */
};
-#ifndef __BYTE_ORDER
-#error "No byte order defined in __BYTE_ORDER"
+#ifdef __BAREBOX__
+# ifdef __LITTLE_ENDIAN
+# define ENVFS_ORDER_LITTLE
+# elif defined __BIG_ENDIAN
+# define ENVFS_ORDER_BIG
+# else
+# error "could not determine byte order"
+# endif
+#else
+# if __BYTE_ORDER == __LITTLE_ENDIAN
+# define ENVFS_ORDER_LITTLE
+# elif __BYTE_ORDER == __BIG_ENDIAN
+# define ENVFS_ORDER_BIG
+# else
+# error "could not determine byte order"
+# endif
#endif
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+#ifdef ENVFS_ORDER_LITTLE
#define ENVFS_16(x) (x)
#define ENVFS_24(x) (x)
#define ENVFS_32(x) (x)
@@ -46,7 +60,7 @@ struct envfs_super {
#define ENVFS_GET_OFFSET(x) ((x)->offset)
#define ENVFS_SET_OFFSET(x,y) ((x)->offset = (y))
#define ENVFS_SET_NAMELEN(x,y) ((x)->namelen = (y))
-#elif __BYTE_ORDER == __BIG_ENDIAN
+#elif defined ENVFS_ORDER_BIG
#ifdef __KERNEL__
#define ENVFS_16(x) swab16(x)
#define ENVFS_24(x) ((swab32(x)) >> 8)
diff --git a/include/linux/byteorder/generic.h b/include/linux/byteorder/generic.h
index aab8f4b..2d68d99 100644
--- a/include/linux/byteorder/generic.h
+++ b/include/linux/byteorder/generic.h
@@ -78,13 +78,6 @@
*
*/
-#ifndef __LITTLE_ENDIAN
-#define __LITTLE_ENDIAN 1234
-#endif
-#ifndef __BIG_ENDIAN
-#define __BIG_ENDIAN 4321
-#endif
-
#if defined(__KERNEL__)
/*
* inside the kernel, we can use nicknames;
diff --git a/include/usb/usb.h b/include/usb/usb.h
index 296e4e8..8691400 100644
--- a/include/usb/usb.h
+++ b/include/usb/usb.h
@@ -270,12 +270,14 @@ void usb_rescan(void);
((x_ & 0xFF000000UL) >> 24)); \
})
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+#ifdef __LITTLE_ENDIAN
# define swap_16(x) (x)
# define swap_32(x) (x)
-#else
+#elif defined BIG_ENDIAN
# define swap_16(x) __swap_16(x)
# define swap_32(x) __swap_32(x)
+#else
+#error "could not determine byte order"
#endif
/*
--
1.7.10
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
next prev parent reply other threads:[~2012-06-28 10:19 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-06-28 9:44 [PATCH 0/4] MIPS: XBurst: fix the JZ4755's clocksource Antony Pavlov
2012-06-28 9:44 ` [PATCH 1/4] byteorder: fix {BIG,LITTLE}_ENDIAN defines Antony Pavlov
2012-06-28 10:19 ` Sascha Hauer [this message]
2012-06-28 9:44 ` [PATCH 2/4] clocksource: move the NSEC_PER_SEC constant to common header Antony Pavlov
2012-06-28 9:44 ` [PATCH 3/4] MIPS: XBurst: fix the JZ4755's clocksource input frequency value Antony Pavlov
2012-06-28 10:33 ` Sascha Hauer
2012-06-28 9:44 ` [PATCH 4/4] MIPS: XBurst: use clocks_calc_mult_shift() for JZ4755's clocksource Antony Pavlov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20120628101917.GU1623@pengutronix.de \
--to=s.hauer@pengutronix.de \
--cc=antonynpavlov@gmail.com \
--cc=barebox@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox