KVM Build error with 2.6.26
Hi,
Against latest kvm commit, 9644a6d164e3d6d0532ddb064393293134f31ab2. KVM compile meet error on 2.6.26.2.
[root@kvm-build kernel]# make
rm -f include/asm
ln -sf asm-x86 include/asm
ln -sf asm-x86 include-compat/asm
make -C /lib/modules/2.6.26.2/build M=pwd \
LINUXINCLUDE="-Ipwd/include -Iinclude
-Iarch/x86/include -I`pwd`/include-compat \
-include include/linux/autoconf.h \
-include `pwd`/x86/external-module-compat.h"make[1]: Entering directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
CC [M] /home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c: In function 'gfn_to_pfn':/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c:742: error:
implicit declaration of function 'get_user_pages_fast' make[3]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o] Error 1 make[2]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86] Error 2 make[1]: *** [_module_/home/build/gitrepo/test/kvm-userspace/kernel] Error 2
make[1]: Leaving directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
Best Regards
Jiajun
Quoted 13 lines · Jiajun
Hi,
Against latest kvm commit,9644a6d164e3d6d0532ddb064393293134f31ab2. KVM compile meet error on 2.6.26.2.[root@kvm-build kernel]# make
rm -finclude/asm
ln -sf asm-x86include/asm
ln -sf asm-x86 include-compat/asm
make -C/lib/modules/2.6.26.2/buildM=pwd\
LINUXINCLUDE="-Ipwd/include -Iinclude-Iarch/x86/include -I`pwd`/include-compat \ -include include/linux/autoconf.h \ -include `pwd`/x86/external-module-compat.h"
Quoted 13 lines
make[1]: Entering directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
CC [M]/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c: In function 'gfn_to_pfn':/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c:742: error:
implicit declaration of function 'get_user_pages_fast' make[3]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o] Error 1 make[2]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86] Error 2 make[1]: *** [_module_/home/build/gitrepo/test/kvm-userspace/kernel] Error 2
make[1]: Leaving directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
I can confirm this bug. Looks like Marcelo's get_user_pages_fast patches lack corresponding compat wrapping (in kvm-userspace)...
Jan
Signature · 2 lines
Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux
PATCH 1/1 add compat wrapper for get_user_pages_fast
Quoted 28 lines · Jan Kiszka
Quoted 13 lines · Jiajun
Hi,
Against latest kvm commit,9644a6d164e3d6d0532ddb064393293134f31ab2. KVM compile meet error on 2.6.26.2.[root@kvm-build kernel]# make
rm -finclude/asm
ln -sf asm-x86include/asm
ln -sf asm-x86 include-compat/asm
make -C/lib/modules/2.6.26.2/buildM=pwd\
LINUXINCLUDE="-Ipwd/include -Iinclude-Iarch/x86/include -I`pwd`/include-compat \ -include include/linux/autoconf.h \ -include `pwd`/x86/external-module-compat.h"Quoted 13 lines
make[1]: Entering directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
CC [M]/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c: In function 'gfn_to_pfn':/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.c:742: error:
implicit declaration of function 'get_user_pages_fast' make[3]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86/kvm_main.o] Error 1 make[2]: *** [/home/build/gitrepo/test/kvm-userspace/kernel/x86] Error 2 make[1]: *** [_module_/home/build/gitrepo/test/kvm-userspace/kernel] Error 2
make[1]: Leaving directory `/usr/src/redhat/BUILD/kernel-2.6.26.2'
Quoted 2 lines
I can confirm this bug. Looks like Marcelo's
get_user_pages_fastpatches lack corresponding compat wrapping (in kvm-userspace)...
Not sure if this is correct, but here is at least a compile fix.
Note that the original mmem_map locking scope was partly far broader on older kernels than with Marcelo's patch and this fix now. Could anyone comment on the correctness?
---------
--- kernel/external-module-compat-comm.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
kernel/external-module-compat-comm.h +17 -0
@@ -531,3 +531,20 @@ struct pci_dev;struct pci_dev *pci_get_bus_and_slot (unsigned int bus,unsigned int devfn);#endif + +#if LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,27 ) + +static inline int get_user_pages_fast (unsigned long start,int nr_pages, +int write,struct page **pages) +{ +int npages; + +down_read (¤t->mm->mmap_sem); + npages =get_user_pages (current, current->mm, start, nr_pages, write, +0 , pages,NULL ); +up_read (¤t->mm->mmap_sem); + +return npages; +} + +#endif
Re: [PATCH] add compat wrapper for get_user_pages_fast
Quoted 1 line · Jan Kiszka
Not sure if this is correct, but here is at least a compile fix.
Looks good.
Quoted 3 lines
Note that the original
mmem_maplocking scope was partly far broader on older kernels than with Marcelo's patch and this fix now. Could anyone comment on the correctness?
I think all cases have been covered (vma manipulation, find_vma callers).
Quoted 33 lines
---------
Signed-off-by Jan Kiszka <[email protected]>--- kernel/external-module-compat-comm.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)Index:
b/kernel/external-module-compat-comm.h
===================================================================kernel/external-module-compat-comm.h +17
-0@@ -531,3 +531,20 @@ struct pci_dev;struct pci_dev *pci_get_bus_and_slot (unsigned int bus,unsigned int devfn);#endif + +#if LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,27 ) + +static inline int get_user_pages_fast (unsigned long start,int nr_pages, +int write,struct page **pages) +{ +int npages; + +down_read (¤t->mm->mmap_sem); + npages =get_user_pages (current, current->mm, start, nr_pages, write, +0 , pages,NULL ); +up_read (¤t->mm->mmap_sem); + +return npages; +} + +#endif
Re: [PATCH] add compat wrapper for get_user_pages_fast
In reply to Jan KiszkaQuoted 3 lines · Jan Kiszka
Not sure if this is correct, but here is at least a compile fix.
I think it is.
Quoted 4 lines
Note that the original
mmem_maplocking scope was partly far broader on older kernels than with Marcelo's patch and this fix now. Could anyone comment on the correctness?
Since get_user_pages_fast() falls back to your code sequence, this is at least as safe as mainline. So I applied it, thanks.
Signature · 2 lines
I have a truly marvellous patch that fixes the bug which this signature is too narrow to contain.