arm,arm64: Conditionalize bio_vec usage
Subsystems (2)
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
Commit 3d1975b57097 (arm,arm64: do not always merge bio_vec if we are running on Xen) unconditionally added code using the bio_vec typedefs which causes build errors on configurations where CONFIG_BLOCK is disabled.
Add #ifdef CONFIG_BLOCK protection to fix this.
--- arch/arm/include/asm/io.h | 2 ++ arch/arm64/include/asm/io.h | 2 ++ 2 files changed, 4 insertions(+)
arch/arm/include/asm/io.h +2 -0
@@ -367,6 +367,7 @@ struct pci_dev;extern void pci_iounmap (struct pci_dev *dev,void __iomem *addr); +#ifdef CONFIG_BLOCK /* * can the hardware map this into one segment or not, given no other * constraints.
@@ -379,6 +380,7 @@ extern bool xen_biovec_phys_mergeable(const struct bio_vec *vec1,#define BIOVEC_PHYS_MERGEABLE (vec1, vec2) \ (__BIOVEC_PHYS_MERGEABLE(vec1, vec2) && \ (!xen_domain() || xen_biovec_phys_mergeable(vec1, vec2))) +#endif #ifdef CONFIG_MMU #define ARCH_HAS_VALID_PHYS_ADDR_RANGE
arch/arm64/include/asm/io.h +2 -0
@@ -266,11 +266,13 @@ extern int devmem_is_allowed(unsigned long pfn);*/ #define xlate_dev_kmem_ptr (p) p +#ifdef CONFIG_BLOCK extern bool xen_biovec_phys_mergeable (const struct bio_vec *vec1,const struct bio_vec *vec2);#define BIOVEC_PHYS_MERGEABLE (vec1, vec2) \ (__BIOVEC_PHYS_MERGEABLE(vec1, vec2) && \ (!xen_domain() || xen_biovec_phys_mergeable(vec1, vec2))) +#endif #endif /* __KERNEL__ */ #endif /* __ASM_IO_H */
Signature · 1 line
1.8.4
Catalin Marinas
maintainerPATCH 1/1 arm,arm64: Conditionalize bio_vec usage
Quoted 4 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.
I guess this commit is only in -next. Would it have the same hash when hitting mainline? Otherwise I'm fine with the patch, maybe Stephano can pick it together with the patch that breaks it.
Signature · 1 line
Catalin
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
Quoted 9 lines · Catalin Marinas
Quoted 4 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.I guess this commit is only in -next. Would it have the same hash when hitting mainline? Otherwise I'm fine with the patch, maybe Stephano can pick it together with the patch that breaks it.
I sent this patch earlier today to fix the problem:
http://marc.info/?l=linux-kernel&m=138374169030300&w=2
I prefer it to adding more ifdefs. I was going to add it to linux-next as soon as possible and then send it upstream together with the rest of the series.
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
Quoted 11 lines · Stefano Stabellini
Quoted 9 lines · Catalin Marinas
Quoted 4 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.I guess this commit is only in -next. Would it have the same hash when hitting mainline? Otherwise I'm fine with the patch, maybe Stephano can pick it together with the patch that breaks it.
Quoted 7 lines
I sent this patch earlier today to fix the problem:
http://marc.info/?l=linux-kernel&m=138374169030300&w=2
I prefer it to adding more ifdefs. I was going to add it to linux-next as soon as possible and then send it upstream together with the rest of the series.
I generally prefer to avoid #ifdefs too. Although there are cases where an #ifdef can be an advantage as well. For instance if you call the xen_biovec_phys_mergeable() function and don't have CONFIG_BLOCK enabled the kernel will happily compile but fail to link. That might be somewhat more difficult to figure out than a compile error. Perhaps it won't.
I have no objections to your proposed patch, and it fixes the build issues, so I'm good.
Thierry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131111/e29912cc/attachment.sig>
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
In reply to Thierry RedingQuoted 8 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.Add #ifdef
CONFIG_BLOCKprotection to fix this.Signed-off-by Thierry Reding <[email protected]>
I commented on the offending patch (which is a good way to make people aware of it instead of posting standalone patches like you did), and Stefano posted patch that declares struct bio_vec; instead.
Either way is ok with me, I'd generally prefer to avoid the ifdefs though.
-Olof
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
Quoted 6 lines · Olof Johansson
Quoted 5 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.
Quoted 6 lines
Quoted 3 lines
Add #ifdef
CONFIG_BLOCKprotection to fix this.Signed-off-by Thierry Reding <[email protected]>I commented on the offending patch (which is a good way to make people aware of it instead of posting standalone patches like you did),
The standalone patch got Stefano's attention, so I don't see why you think it was a bad idea.
I don't expect such patches to always be applied as is. If they aren't proper fixes and someone comes up with a much better way (which they did in this case), then I'm just as happy. Also it actually takes about the same amount of time to write up a patch, compile-test it and send it out than it takes to complain about the breakage by mail. I also find it slightly more helpful than just telling somebody that their patch broke something.
Perhaps this is a bad example because it's really a trivial issue, but if it were anything slightly more complex, then sending a patch for it may save some maintainer the additional work. I don't immediately see what's wrong with that. Perhaps you'd care to elaborate.
Quoted 3 lines
and Stefano posted patch that declares struct
bio_vec; instead.Either way is ok with me, I'd generally prefer to avoid the ifdefs though.
I agree, that's much nicer than the #ifdef.
Thierry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131111/1700a2d2/attachment.sig>
PATCH 1/1 arm,arm64: Conditionalize bio_vec usage
In reply to Thierry RedingQuoted 8 lines · Thierry Reding
Commit
3d1975b57097(arm,arm64: do not always mergebio_vecif we are running on Xen) unconditionally added code using thebio_vectypedefs which causes build errors on configurations whereCONFIG_BLOCKis disabled.Add #ifdef
CONFIG_BLOCKprotection to fix this.Signed-off-by Thierry Reding <[email protected]>
Thierry, thank you for the patch.
However I was thinking of declaring struct bio_vec instead, see:
http://marc.info/?l=linux-kernel&m=138374169030300&w=2
Quoted 45 lines
arch/arm/include/asm/io.h | 2 ++ arch/arm64/include/asm/io.h | 2 ++ 2 files changed, 4 insertions(+)arch/arm/include/asm/io.h +2
-0@@ -367,6 +367,7 @@ struct pci_dev;extern void pci_iounmap (struct pci_dev *dev,void __iomem *addr); +#ifdef CONFIG_BLOCK /* * can the hardware map this into one segment or not, given no other * constraints. @@ -379,6 +380,7 @@ extern bool xen_biovec_phys_mergeable(const struct bio_vec *vec1,#define BIOVEC_PHYS_MERGEABLE (vec1, vec2) \ (__BIOVEC_PHYS_MERGEABLE(vec1, vec2) && \ (!xen_domain() || xen_biovec_phys_mergeable(vec1, vec2))) +#endif #ifdef CONFIG_MMU #define ARCH_HAS_VALID_PHYS_ADDR_RANGE arch/arm64/include/asm/io.h +2
-0@@ -266,11 +266,13 @@ extern int devmem_is_allowed(unsigned long pfn);*/ #define xlate_dev_kmem_ptr (p) p +#ifdef CONFIG_BLOCK extern bool xen_biovec_phys_mergeable (const struct bio_vec *vec1,const struct bio_vec *vec2);#define BIOVEC_PHYS_MERGEABLE (vec1, vec2) \ (__BIOVEC_PHYS_MERGEABLE(vec1, vec2) && \ (!xen_domain() || xen_biovec_phys_mergeable(vec1, vec2))) +#endif #endif /* __KERNEL__ */ #endif /* __ASM_IO_H */ Signature · 1 line
1.8.4