Skip to contentSkip to conversation

build fix: xen on NetBSD/amd64

PATCH In review Christoph Egger 3 participants 11 messages Jun 30, 2009 – Jul 17, 2009

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 build fix: xen on NetBSD/amd64

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
hw/xen_blkif.h +4 -4
index 738b8fe..bb4ff02 100644
@@ -16,8 +16,9 @@ struct blkif_common_response {
 	char dummy;
 };
 
+/* XXX move this into a proper header */
+#define __aligned(x)	__attribute__((__aligned__(x)))
 /* i386 protocol version */
-#pragma pack(push, 4)
 struct blkif_x86_32_request {
 	uint8_t        operation;    /* BLKIF_OP_???                         */
 	uint8_t        nr_segments;  /* number of segments                   */
@@ -25,15 +26,14 @@ struct blkif_x86_32_request {
 	uint64_t       id;           /* private guest value, echoed in resp  */
 	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
-};
+} __aligned(4);
 struct blkif_x86_32_response {
 	uint64_t        id;              /* copied from request */
 	uint8_t         operation;       /* copied from request */
 	int16_t         status;          /* BLKIF_RSP_???       */
-};
+} __aligned(4);
 typedef struct blkif_x86_32_request blkif_x86_32_request_t;
 typedef struct blkif_x86_32_response blkif_x86_32_response_t;
-#pragma pack(pop)
 
 /* x86_64 protocol version */
 struct blkif_x86_64_request {
#1 lore raw
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

There really isn't pragma pack on NetBSD? That's ashame.

Regards,

Anthony Liguori

#2 lore raw
Quoted 10 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 4 lines
Quoted 1 line

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

#if __GNUC_PREREQ__(2, 7)

#define __packed        __attribute__((__packed__))
#define __aligned(x)    __attribute__((__aligned__(x)))
#define __section(x)    __attribute__((__section__(x)))
#elif defined(__PCC__)
#define __packed        _Pragma("packed 1")
#define __aligned(x)    _Pragma("aligned " __STRING(x))
#define __section(x)    _Pragma("section " ## x)
#elif defined(__lint__)
#define __packed        /* delete */
#define __aligned(x)    /* delete */
#define __section(x)    /* delete */
#else
#define __packed        error: no __packed for this compiler
#define __aligned(x)    error: no __aligned for this compiler
#define __section(x)    error: no __section for this compiler

#endif

Quoted 1 line

There really isn't pragma pack on NetBSD? That's ashame.

No. Above defines are sufficient and portable.
Now you know where the __aligned define in my patch comes from. :-)

Christoph

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
#3 lore raw
Quoted 34 lines · Christoph Egger
Quoted 10 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 4 lines
Quoted 1 line

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

#if __GNUC_PREREQ__(2, 7)

#define __packed        __attribute__((__packed__))
#define __aligned(x)    __attribute__((__aligned__(x)))
#define __section(x)    __attribute__((__section__(x)))
#elif defined(__PCC__)
#define __packed        _Pragma("packed 1")
#define __aligned(x)    _Pragma("aligned " __STRING(x))
#define __section(x)    _Pragma("section " ## x)
#elif defined(__lint__)
#define __packed        /* delete */
#define __aligned(x)    /* delete */
#define __section(x)    /* delete */
#else
#define __packed        error: no __packed for this compiler
#define __aligned(x)    error: no __aligned for this compiler
#define __section(x)    error: no __section for this compiler
Quoted 6 lines

#endif

Quoted 1 line

There really isn't pragma pack on NetBSD? That's ashame.

No. Above defines are sufficient and portable.

Above defines are sufficient but not anywhere near portable.

Quoted 1 line

Now you know where the __aligned define in my patch comes from. :-)

You should conditionally include include sys/cdefs.h on NetBSD and not defining your own version of __aligned (which is forbidden by the standard).

Few more points:

a. the file in question already uses aligned attribute _inside_ the
structures without any macros

b. unlike all other xen* files the header guard is __XEN_BLKIF_H__
and not QEMU_XEN_BLKIF_H (that should be fixed too)

Signature · 1 line
mailto:[email protected]
#4 lore raw
Quoted 37 lines · On Friday 10 July 2009 12:18:39 malc wrote:
Quoted 36 lines · Christoph Egger
Quoted 10 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 4 lines
Quoted 1 line

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

#if __GNUC_PREREQ__(2, 7)

#define __packed        __attribute__((__packed__))
#define __aligned(x)    __attribute__((__aligned__(x)))
#define __section(x)    __attribute__((__section__(x)))
#elif defined(__PCC__)
#define __packed        _Pragma("packed 1")
#define __aligned(x)    _Pragma("aligned " __STRING(x))
#define __section(x)    _Pragma("section " ## x)
#elif defined(__lint__)
#define __packed        /* delete */
#define __aligned(x)    /* delete */
#define __section(x)    /* delete */
#else
#define __packed        error: no __packed for this compiler
#define __aligned(x)    error: no __aligned for this compiler
#define __section(x)    error: no __section for this compiler

#endif

Quoted 5 lines
Quoted 3 lines
Quoted 1 line

There really isn't pragma pack on NetBSD? That's ashame.

No. Above defines are sufficient and portable.

Above defines are sufficient but not anywhere near portable.

Well, across different compilers, they are.

Quoted 5 lines
Quoted 1 line

Now you know where the __aligned define in my patch comes from. :-)

You should conditionally include include sys/cdefs.h on NetBSD and not defining your own version of __aligned (which is forbidden by the standard).

sys/cdefs.h is included implicitely by any libc standard header on NetBSD. No need for an #ifdef __NetBSD__ ... #endif for this.

Quoted 4 lines

Few more points:

a. the file in question already uses aligned attribute _inside_ the
structures without any macros

So the pragmas have no effect then ? If so, then just remove them.

Quoted 2 lines

b. unlike all other xen* files the header guard is __XEN_BLKIF_H__
and not QEMU_XEN_BLKIF_H (that should be fixed too)

Easy to fix. Will you do or should I send a diff ?

Christoph

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
#5 lore raw
Quoted 38 lines · Christoph Egger
Quoted 37 lines · On Friday 10 July 2009 12:18:39 malc wrote:
Quoted 36 lines · Christoph Egger
Quoted 10 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 4 lines
Quoted 1 line

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

#if __GNUC_PREREQ__(2, 7)

#define __packed        __attribute__((__packed__))
#define __aligned(x)    __attribute__((__aligned__(x)))
#define __section(x)    __attribute__((__section__(x)))
#elif defined(__PCC__)
#define __packed        _Pragma("packed 1")
#define __aligned(x)    _Pragma("aligned " __STRING(x))
#define __section(x)    _Pragma("section " ## x)
#elif defined(__lint__)
#define __packed        /* delete */
#define __aligned(x)    /* delete */
#define __section(x)    /* delete */
#else
#define __packed        error: no __packed for this compiler
#define __aligned(x)    error: no __aligned for this compiler
#define __section(x)    error: no __section for this compiler

#endif

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

There really isn't pragma pack on NetBSD? That's ashame.

No. Above defines are sufficient and portable.

Above defines are sufficient but not anywhere near portable.

Well, across different compilers, they are.

What i meant is that they (the system provider) are free to do that, you (the user) can not do that though.

Quoted 9 lines
Quoted 6 lines
Quoted 2 lines

Now you know where the __aligned define in my patch comes from. :-)

You should conditionally include include sys/cdefs.h on NetBSD and not defining your own version of __aligned (which is forbidden by the standard).

sys/cdefs.h is included implicitely by any libc standard header on NetBSD. No need for an #ifdef __NetBSD__ ... #endif for this.

Then i actually fail to see how your original patch is supposed to work at all, gcc should raise hell about redefinition of __aligned.

Quoted 6 lines
Quoted 4 lines

Few more points:

a. the file in question already uses aligned attribute _inside_ the
structures without any macros

So the pragmas have no effect then ? If so, then just remove them.

Not really, the individual members are marked with attribute aligned (just take a look), not the whole structures.

Quoted 6 lines
Quoted 3 lines

b. unlike all other xen* files the header guard is __XEN_BLKIF_H__
and not QEMU_XEN_BLKIF_H (that should be fixed too)

Easy to fix. Will you do or should I send a diff ?

That would be very nice of you.

Signature · 1 line
mailto:[email protected]
#6 lore raw
Quoted 72 lines · On Friday 10 July 2009 17:22:04 malc wrote:
Quoted 38 lines · Christoph Egger
Quoted 37 lines · On Friday 10 July 2009 12:18:39 malc wrote:
Quoted 36 lines · Christoph Egger
Quoted 10 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 4 lines
Quoted 1 line

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

#if __GNUC_PREREQ__(2, 7)

#define __packed        __attribute__((__packed__))
#define __aligned(x)    __attribute__((__aligned__(x)))
#define __section(x)    __attribute__((__section__(x)))
#elif defined(__PCC__)
#define __packed        _Pragma("packed 1")
#define __aligned(x)    _Pragma("aligned " __STRING(x))
#define __section(x)    _Pragma("section " ## x)
#elif defined(__lint__)
#define __packed        /* delete */
#define __aligned(x)    /* delete */
#define __section(x)    /* delete */
#else
#define __packed        error: no __packed for this compiler
#define __aligned(x)    error: no __aligned for this compiler
#define __section(x)    error: no __section for this compiler

#endif

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

There really isn't pragma pack on NetBSD? That's ashame.

No. Above defines are sufficient and portable.

Above defines are sufficient but not anywhere near portable.

Well, across different compilers, they are.

What i meant is that they (the system provider) are free to do that, you (the user) can not do that though.

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

Now you know where the __aligned define in my patch comes from. :-)

You should conditionally include include sys/cdefs.h on NetBSD and not defining your own version of __aligned (which is forbidden by the standard).

sys/cdefs.h is included implicitely by any libc standard header on NetBSD. No need for an #ifdef __NetBSD__ ... #endif for this.

Then i actually fail to see how your original patch is supposed to work at all, gcc should raise hell about redefinition of __aligned.

Quoted 6 lines
Quoted 4 lines

Few more points:

a. the file in question already uses aligned attribute _inside_ the
structures without any macros

So the pragmas have no effect then ? If so, then just remove them.

Not really, the individual members are marked with attribute aligned (just take a look), not the whole structures.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

b. unlike all other xen* files the header guard is __XEN_BLKIF_H__
and not QEMU_XEN_BLKIF_H (that should be fixed too)

Easy to fix. Will you do or should I send a diff ?

That would be very nice of you.

Is attached patch fine with you ?

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
hw/xen_blkif.h +3 -5
index 738b8fe..f47f9a4 100644
@@ -1,5 +1,5 @@
-#ifndef __XEN_BLKIF_H__
-#define __XEN_BLKIF_H__
+#ifndef __QEMU_BLKIF_H__
+#define __QEMU_BLKIF_H__
 
 #include <xen/io/ring.h>
 #include <xen/io/blkif.h>
@@ -17,7 +17,6 @@ struct blkif_common_response {
 };
 
 /* i386 protocol version */
-#pragma pack(push, 4)
 struct blkif_x86_32_request {
 	uint8_t        operation;    /* BLKIF_OP_???                         */
 	uint8_t        nr_segments;  /* number of segments                   */
@@ -33,7 +32,6 @@ struct blkif_x86_32_response {
 };
 typedef struct blkif_x86_32_request blkif_x86_32_request_t;
 typedef struct blkif_x86_32_response blkif_x86_32_response_t;
-#pragma pack(pop)
 
 /* x86_64 protocol version */
 struct blkif_x86_64_request {
@@ -100,4 +98,4 @@ static void inline blkif_get_x86_64_req(blkif_request_t *dst, blkif_x86_64_reque
 		dst->seg[i] = src->seg[i];
 }
 
-#endif /* __XEN_BLKIF_H__ */
+#endif /* __QEMU_BLKIF_H__ */
#7 lore raw
Quoted 20 lines · Christoph Egger
Quoted 19 lines · On Friday 10 July 2009 17:22:04 malc wrote:
Quoted 18 lines · Christoph Egger
Quoted 17 lines · On Friday 10 July 2009 12:18:39 malc wrote:
Quoted 16 lines · Christoph Egger
Quoted 14 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 10 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

[..snip..]

Quoted 7 lines
Quoted 2 lines

That would be very nice of you.

Is attached patch fine with you ?

No. It replaces one identifier starting with __ with another.

Signature · 1 line
mailto:[email protected]
#8 lore raw
In reply to Christoph Egger
Quoted 13 lines · Christoph Egger
Quoted 11 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 9 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

NetBSD is allowed to use these names in system headers but according to the C standard, normal programs like QEMU aren't. So we need to use something other than __aligned.

Regards,

Anthony Liguori

#9 lore raw
Quoted 14 lines · On Friday 10 July 2009 14:58:48 Anthony Liguori wrote:
Quoted 13 lines · Christoph Egger
Quoted 12 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 10 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

NetBSD is allowed to use these names in system headers but according to the C standard, normal programs like QEMU aren't. So we need to use something other than __aligned.

What about __qemu_aligned ?

#ifdef __NetBSD__
#define __qemu_aligned(x) __aligned(x)
#else
#define __qemu_aligned(x) __attribute__((__aligned__(x)))
#endif

Quoted 3 lines

Regards,

Anthony Liguori

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
#10 lore raw
Quoted 24 lines · Christoph Egger
Quoted 14 lines · On Friday 10 July 2009 14:58:48 Anthony Liguori wrote:
Quoted 13 lines · Christoph Egger
Quoted 12 lines · On Thursday 09 July 2009 21:21:06 Anthony Liguori wrote:
Quoted 10 lines · Christoph Egger

Hi!

Attached patch fixes this build error on NetBSD/amd64:

hw/xen_blkif.h:20: warning: #pragma pack(psuh[, id], <n>) is not supported on this target
hw/xen_blkif.h:36: warning: #pragma pack(pop[, id], <n>) is not supported on this target

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

You'll invoke the fury of malc for introducing an identifier that begins with '__' :-)

In NetBSD, there is this in <sys/cdefs.h> :

NetBSD is allowed to use these names in system headers but according to the C standard, normal programs like QEMU aren't. So we need to use something other than __aligned.

Quoted 7 lines

What about __qemu_aligned ?

#ifdef __NetBSD__
#define __qemu_aligned(x) __aligned(x)
#else
#define __qemu_aligned(x) __attribute__((__aligned__(x)))
#endif

Let me quote the standard once again:

7.1.3 Reserved identifiers
[..snip..]
-- All identifiers that begin with an underscore and
either an uppercase letter or another underscore are always reserved for any use.

Signature · 1 line
mailto:[email protected]