| Message ID | 20220802164134.1851910-1-berrange@redhat.com |
|---|---|
| State | New |
| Headers | show |
| Series | [for,7.1] linux-user: fix compat with glibc >= 2.36 sys/mount.h | expand |
On Tue, Aug 02, 2022 at 12:41:34PM -0400, Daniel P. Berrangé wrote: > The latest glibc 2.36 has extended sys/mount.h so that it > defines the FSCONFIG_* enum constants. These are historically > defined in linux/mount.h, and thus if you include both headers > the compiler complains: > > In file included from /usr/include/linux/fs.h:19, > from ../linux-user/syscall.c:98: > /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' > 95 | enum fsconfig_command { > | ^~~~~~~~~~~~~~~~ > In file included from ../linux-user/syscall.c:31: > /usr/include/sys/mount.h:189:6: note: originally defined here > 189 | enum fsconfig_command > | ^~~~~~~~~~~~~~~~ > /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' > 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' > 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > ...snip... > > QEMU doesn't include linux/mount.h, but it does use > linux/fs.h and thus gets linux/mount.h indirectly. > > glibc acknowledges this problem but does not appear to > be intending to fix it in the forseeable future, simply > documenting it as a known incompatibility with no > workaround: > > https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E > https://sourceware.org/glibc/wiki/Synchronizing_Headers > > To address this requires either removing use of sys/mount.h > or linux/fs.h, despite QEMU needing declarations from > both. > > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. > > Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> > --- > linux-user/syscall.c | 18 ++++++++++++++++++ > meson.build | 2 ++ > 2 files changed, 20 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index b27a6552aa..52d178afe7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -95,7 +95,25 @@ > #include <linux/soundcard.h> > #include <linux/kd.h> > #include <linux/mtio.h> > + > +#ifdef HAVE_SYS_MOUNT_FSCONFIG > +/* > + * glibc >= 2.36 linux/mount.h conflicts with sys/mount.h, > + * which in turn prevents use of linux/fs.h. So we have to > + * define the constants ourselves for now. > + */ > +#define FS_IOC_GETFLAGS _IOR('f', 1, long) > +#define FS_IOC_SETFLAGS _IOW('f', 2, long) > +#define FS_IOC_GETVERSION _IOR('v', 1, long) > +#define FS_IOC_SETVERSION _IOW('v', 2, long) > +#define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) > +#define FS_IOC32_GETFLAGS _IOR('f', 1, int) > +#define FS_IOC32_SETFLAGS _IOW('f', 2, int) > +#define FS_IOC32_GETVERSION _IOR('v', 1, int) > +#define FS_IOC32_SETVERSION _IOW('v', 2, int) > +#else > #include <linux/fs.h> > +#endif > #include <linux/fd.h> > #if defined(CONFIG_FIEMAP) > #include <linux/fiemap.h> > diff --git a/meson.build b/meson.build > index 294e9a8f32..30a380752c 100644 > --- a/meson.build > +++ b/meson.build > @@ -1963,6 +1963,8 @@ config_host_data.set('HAVE_OPTRESET', > cc.has_header_symbol('getopt.h', 'optreset')) > config_host_data.set('HAVE_IPPROTO_MPTCP', > cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP')) > +config_host_data.set('HAVE_SYS_MOUNT_FSCONFIG', > + cc.has_header_symbol('sys/mount.h', 'FSCONFIG_SET_FLAG')) > > # has_member > config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID', Dan, which Fedora glibc package shows this problem? I have glibc-2.35.9000-31.fc37.x86_64 and qemu compiled fine. (Also nbdkit which includes linux/fs.h) I see various other build failures in Koji, but not this one as far as I can tell: https://koji.fedoraproject.org/koji/packageinfo?packageID=3685 Rich.
On Tue, Aug 02, 2022 at 07:29:29PM +0100, Richard W.M. Jones wrote: > Dan, which Fedora glibc package shows this problem? I have > glibc-2.35.9000-31.fc37.x86_64 and qemu compiled fine. (Also nbdkit > which includes linux/fs.h) It would help if I enabled a *-linux-user target ... Yes, I can reproduce this now and the patch fixes it, so: Tested-by: Richard W.M. Jones <rjones@redhat.com> Rich.
On Tue, 2 Aug 2022 at 17:43, Daniel P. Berrangé <berrange@redhat.com> wrote: > > The latest glibc 2.36 has extended sys/mount.h so that it > defines the FSCONFIG_* enum constants. These are historically > defined in linux/mount.h, and thus if you include both headers > the compiler complains: > > In file included from /usr/include/linux/fs.h:19, > from ../linux-user/syscall.c:98: > /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' > 95 | enum fsconfig_command { > | ^~~~~~~~~~~~~~~~ > In file included from ../linux-user/syscall.c:31: > /usr/include/sys/mount.h:189:6: note: originally defined here > 189 | enum fsconfig_command > | ^~~~~~~~~~~~~~~~ > /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' > 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' > 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > ...snip... > > QEMU doesn't include linux/mount.h, but it does use > linux/fs.h and thus gets linux/mount.h indirectly. > > glibc acknowledges this problem but does not appear to > be intending to fix it in the forseeable future, simply > documenting it as a known incompatibility with no > workaround: > > https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E > https://sourceware.org/glibc/wiki/Synchronizing_Headers > > To address this requires either removing use of sys/mount.h > or linux/fs.h, despite QEMU needing declarations from > both. > > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. I don't suppose anybody has useful contacts with the glibc developers so we can ask them if they can try to maybe not break applications like this, or at least if they do to provide workarounds ? Having to hand-define a bunch of ioctl constants is particularly messy. thanks -- PMM
CC'ing Florian to get the POV of a glibc maintainer on what we've had to do to work around this compatibility brekage. On Tue, Aug 02, 2022 at 12:41:34PM -0400, Daniel P. Berrangé wrote: > The latest glibc 2.36 has extended sys/mount.h so that it > defines the FSCONFIG_* enum constants. These are historically > defined in linux/mount.h, and thus if you include both headers > the compiler complains: > > In file included from /usr/include/linux/fs.h:19, > from ../linux-user/syscall.c:98: > /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' > 95 | enum fsconfig_command { > | ^~~~~~~~~~~~~~~~ > In file included from ../linux-user/syscall.c:31: > /usr/include/sys/mount.h:189:6: note: originally defined here > 189 | enum fsconfig_command > | ^~~~~~~~~~~~~~~~ > /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' > 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' > 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > ...snip... > > QEMU doesn't include linux/mount.h, but it does use > linux/fs.h and thus gets linux/mount.h indirectly. > > glibc acknowledges this problem but does not appear to > be intending to fix it in the forseeable future, simply > documenting it as a known incompatibility with no > workaround: > > https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E > https://sourceware.org/glibc/wiki/Synchronizing_Headers > > To address this requires either removing use of sys/mount.h > or linux/fs.h, despite QEMU needing declarations from > both. > > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. > > Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> > --- > linux-user/syscall.c | 18 ++++++++++++++++++ > meson.build | 2 ++ > 2 files changed, 20 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index b27a6552aa..52d178afe7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -95,7 +95,25 @@ > #include <linux/soundcard.h> > #include <linux/kd.h> > #include <linux/mtio.h> > + > +#ifdef HAVE_SYS_MOUNT_FSCONFIG > +/* > + * glibc >= 2.36 linux/mount.h conflicts with sys/mount.h, > + * which in turn prevents use of linux/fs.h. So we have to > + * define the constants ourselves for now. > + */ > +#define FS_IOC_GETFLAGS _IOR('f', 1, long) > +#define FS_IOC_SETFLAGS _IOW('f', 2, long) > +#define FS_IOC_GETVERSION _IOR('v', 1, long) > +#define FS_IOC_SETVERSION _IOW('v', 2, long) > +#define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) > +#define FS_IOC32_GETFLAGS _IOR('f', 1, int) > +#define FS_IOC32_SETFLAGS _IOW('f', 2, int) > +#define FS_IOC32_GETVERSION _IOR('v', 1, int) > +#define FS_IOC32_SETVERSION _IOW('v', 2, int) > +#else > #include <linux/fs.h> > +#endif > #include <linux/fd.h> > #if defined(CONFIG_FIEMAP) > #include <linux/fiemap.h> > diff --git a/meson.build b/meson.build > index 294e9a8f32..30a380752c 100644 > --- a/meson.build > +++ b/meson.build > @@ -1963,6 +1963,8 @@ config_host_data.set('HAVE_OPTRESET', > cc.has_header_symbol('getopt.h', 'optreset')) > config_host_data.set('HAVE_IPPROTO_MPTCP', > cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP')) > +config_host_data.set('HAVE_SYS_MOUNT_FSCONFIG', > + cc.has_header_symbol('sys/mount.h', 'FSCONFIG_SET_FLAG')) > > # has_member > config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID', > -- > 2.37.1 > With regards, Daniel
* Daniel P. Berrangé: > CC'ing Florian to get the POV of a glibc maintainer on what we've > had to do to work around this compatibility brekage. Meh, that's really not great. 8-( I'll see if I can add a similar workaround to the glibc headers. Unfortunately I'm busy with other stuff right now, so it's going to be a few weeks until I can work on this (probably). Thanks, Florian
Laurent: ping, Can you consider queuing this so 7.1 isn't broken with latest glibc releases. On Tue, Aug 02, 2022 at 12:41:34PM -0400, Daniel P. Berrangé wrote: > The latest glibc 2.36 has extended sys/mount.h so that it > defines the FSCONFIG_* enum constants. These are historically > defined in linux/mount.h, and thus if you include both headers > the compiler complains: > > In file included from /usr/include/linux/fs.h:19, > from ../linux-user/syscall.c:98: > /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' > 95 | enum fsconfig_command { > | ^~~~~~~~~~~~~~~~ > In file included from ../linux-user/syscall.c:31: > /usr/include/sys/mount.h:189:6: note: originally defined here > 189 | enum fsconfig_command > | ^~~~~~~~~~~~~~~~ > /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' > 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' > 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > ...snip... > > QEMU doesn't include linux/mount.h, but it does use > linux/fs.h and thus gets linux/mount.h indirectly. > > glibc acknowledges this problem but does not appear to > be intending to fix it in the forseeable future, simply > documenting it as a known incompatibility with no > workaround: > > https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E > https://sourceware.org/glibc/wiki/Synchronizing_Headers > > To address this requires either removing use of sys/mount.h > or linux/fs.h, despite QEMU needing declarations from > both. > > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. > > Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> > --- > linux-user/syscall.c | 18 ++++++++++++++++++ > meson.build | 2 ++ > 2 files changed, 20 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index b27a6552aa..52d178afe7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -95,7 +95,25 @@ > #include <linux/soundcard.h> > #include <linux/kd.h> > #include <linux/mtio.h> > + > +#ifdef HAVE_SYS_MOUNT_FSCONFIG > +/* > + * glibc >= 2.36 linux/mount.h conflicts with sys/mount.h, > + * which in turn prevents use of linux/fs.h. So we have to > + * define the constants ourselves for now. > + */ > +#define FS_IOC_GETFLAGS _IOR('f', 1, long) > +#define FS_IOC_SETFLAGS _IOW('f', 2, long) > +#define FS_IOC_GETVERSION _IOR('v', 1, long) > +#define FS_IOC_SETVERSION _IOW('v', 2, long) > +#define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) > +#define FS_IOC32_GETFLAGS _IOR('f', 1, int) > +#define FS_IOC32_SETFLAGS _IOW('f', 2, int) > +#define FS_IOC32_GETVERSION _IOR('v', 1, int) > +#define FS_IOC32_SETVERSION _IOW('v', 2, int) > +#else > #include <linux/fs.h> > +#endif > #include <linux/fd.h> > #if defined(CONFIG_FIEMAP) > #include <linux/fiemap.h> > diff --git a/meson.build b/meson.build > index 294e9a8f32..30a380752c 100644 > --- a/meson.build > +++ b/meson.build > @@ -1963,6 +1963,8 @@ config_host_data.set('HAVE_OPTRESET', > cc.has_header_symbol('getopt.h', 'optreset')) > config_host_data.set('HAVE_IPPROTO_MPTCP', > cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP')) > +config_host_data.set('HAVE_SYS_MOUNT_FSCONFIG', > + cc.has_header_symbol('sys/mount.h', 'FSCONFIG_SET_FLAG')) > > # has_member > config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID', > -- > 2.37.1 > With regards, Daniel
Le 02/08/2022 à 18:41, Daniel P. Berrangé a écrit : > The latest glibc 2.36 has extended sys/mount.h so that it > defines the FSCONFIG_* enum constants. These are historically > defined in linux/mount.h, and thus if you include both headers > the compiler complains: > > In file included from /usr/include/linux/fs.h:19, > from ../linux-user/syscall.c:98: > /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' > 95 | enum fsconfig_command { > | ^~~~~~~~~~~~~~~~ > In file included from ../linux-user/syscall.c:31: > /usr/include/sys/mount.h:189:6: note: originally defined here > 189 | enum fsconfig_command > | ^~~~~~~~~~~~~~~~ > /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' > 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' > 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ > | ^~~~~~~~~~~~~~~~~ > ...snip... > > QEMU doesn't include linux/mount.h, but it does use > linux/fs.h and thus gets linux/mount.h indirectly. > > glibc acknowledges this problem but does not appear to > be intending to fix it in the forseeable future, simply > documenting it as a known incompatibility with no > workaround: > > https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E > https://sourceware.org/glibc/wiki/Synchronizing_Headers > > To address this requires either removing use of sys/mount.h > or linux/fs.h, despite QEMU needing declarations from > both. > > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. > > Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> > --- > linux-user/syscall.c | 18 ++++++++++++++++++ > meson.build | 2 ++ > 2 files changed, 20 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index b27a6552aa..52d178afe7 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -95,7 +95,25 @@ > #include <linux/soundcard.h> > #include <linux/kd.h> > #include <linux/mtio.h> > + > +#ifdef HAVE_SYS_MOUNT_FSCONFIG > +/* > + * glibc >= 2.36 linux/mount.h conflicts with sys/mount.h, > + * which in turn prevents use of linux/fs.h. So we have to > + * define the constants ourselves for now. > + */ > +#define FS_IOC_GETFLAGS _IOR('f', 1, long) > +#define FS_IOC_SETFLAGS _IOW('f', 2, long) > +#define FS_IOC_GETVERSION _IOR('v', 1, long) > +#define FS_IOC_SETVERSION _IOW('v', 2, long) > +#define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) > +#define FS_IOC32_GETFLAGS _IOR('f', 1, int) > +#define FS_IOC32_SETFLAGS _IOW('f', 2, int) > +#define FS_IOC32_GETVERSION _IOR('v', 1, int) > +#define FS_IOC32_SETVERSION _IOW('v', 2, int) > +#else > #include <linux/fs.h> > +#endif > #include <linux/fd.h> > #if defined(CONFIG_FIEMAP) > #include <linux/fiemap.h> > diff --git a/meson.build b/meson.build > index 294e9a8f32..30a380752c 100644 > --- a/meson.build > +++ b/meson.build > @@ -1963,6 +1963,8 @@ config_host_data.set('HAVE_OPTRESET', > cc.has_header_symbol('getopt.h', 'optreset')) > config_host_data.set('HAVE_IPPROTO_MPTCP', > cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP')) > +config_host_data.set('HAVE_SYS_MOUNT_FSCONFIG', > + cc.has_header_symbol('sys/mount.h', 'FSCONFIG_SET_FLAG')) > > # has_member > config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID', Applied to my linux-user-for-7.1 branch. Thanks, Laurent
On Aug 02 2022, Daniel P. Berrangé wrote: > This patch removes linux/fs.h, meaning we have to define > various FS_IOC constants that are now unavailable. This breaks a lot of ioctl emulations, as it lacks their definitions: #define BLKGETSIZE64 _IOR(0x12,114,size_t) #define BLKDISCARD _IO(0x12,119) #define BLKIOMIN _IO(0x12,120) #define BLKIOOPT _IO(0x12,121) #define BLKALIGNOFF _IO(0x12,122) #define BLKPBSZGET _IO(0x12,123) #define BLKDISCARDZEROES _IO(0x12,124) #define BLKSECDISCARD _IO(0x12,125) #define BLKROTATIONAL _IO(0x12,126) #define BLKZEROOUT _IO(0x12,127) #define FIBMAP _IO(0x00,1) #define FICLONE _IOW(0x94, 9, int) #define FIGETBSZ _IO(0x00,2)
On Sat, Oct 01, 2022 at 11:34:20PM +0200, Andreas Schwab wrote: > On Aug 02 2022, Daniel P. Berrangé wrote: > > > This patch removes linux/fs.h, meaning we have to define > > various FS_IOC constants that are now unavailable. > > This breaks a lot of ioctl emulations, as it lacks their definitions: Urgh, very sorry about that. I of course tested it compiled, but not realizing a bunch of ioctls were #ifdef'd, so we ended up silently disabling them. > > #define BLKGETSIZE64 _IOR(0x12,114,size_t) > #define BLKDISCARD _IO(0x12,119) > #define BLKIOMIN _IO(0x12,120) > #define BLKIOOPT _IO(0x12,121) > #define BLKALIGNOFF _IO(0x12,122) > #define BLKPBSZGET _IO(0x12,123) > #define BLKDISCARDZEROES _IO(0x12,124) > #define BLKSECDISCARD _IO(0x12,125) > #define BLKROTATIONAL _IO(0x12,126) > #define BLKZEROOUT _IO(0x12,127) > > #define FIBMAP _IO(0x00,1) > #define FICLONE _IOW(0x94, 9, int) > #define FIGETBSZ _IO(0x00,2) With the exception of FICLONE from 2015, all the others here have exited in Linux header since 2012. Given that length of time there's no reason for us to conditionalize their usage. We want to see failure if they're unexpectedly missing from headers. I'll prepare a patch to fix all this. With regards, Daniel
On Mon, Oct 03, 2022 at 09:31:52AM +0100, Daniel P. Berrangé wrote: > On Sat, Oct 01, 2022 at 11:34:20PM +0200, Andreas Schwab wrote: > > On Aug 02 2022, Daniel P. Berrangé wrote: > > > > > This patch removes linux/fs.h, meaning we have to define > > > various FS_IOC constants that are now unavailable. > > > > This breaks a lot of ioctl emulations, as it lacks their definitions: > > Urgh, very sorry about that. I of course tested it compiled, but not > realizing a bunch of ioctls were #ifdef'd, so we ended up silently > disabling them. > > > > > #define BLKGETSIZE64 _IOR(0x12,114,size_t) > > #define BLKDISCARD _IO(0x12,119) > > #define BLKIOMIN _IO(0x12,120) > > #define BLKIOOPT _IO(0x12,121) > > #define BLKALIGNOFF _IO(0x12,122) > > #define BLKPBSZGET _IO(0x12,123) > > #define BLKDISCARDZEROES _IO(0x12,124) > > #define BLKSECDISCARD _IO(0x12,125) > > #define BLKROTATIONAL _IO(0x12,126) > > #define BLKZEROOUT _IO(0x12,127) > > > > #define FIBMAP _IO(0x00,1) > > #define FICLONE _IOW(0x94, 9, int) > > #define FIGETBSZ _IO(0x00,2) > > With the exception of FICLONE from 2015, all the others here have exited > in Linux header since 2012. Given that length of time there's no reason > for us to conditionalize their usage. We want to see failure if they're > unexpectedly missing from headers. > > I'll prepare a patch to fix all this. I have a patch series here: https://lists.gnu.org/archive/html/qemu-devel/2022-10/msg00351.html With regards, Daniel
diff --git a/linux-user/syscall.c b/linux-user/syscall.c index b27a6552aa..52d178afe7 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -95,7 +95,25 @@ #include <linux/soundcard.h> #include <linux/kd.h> #include <linux/mtio.h> + +#ifdef HAVE_SYS_MOUNT_FSCONFIG +/* + * glibc >= 2.36 linux/mount.h conflicts with sys/mount.h, + * which in turn prevents use of linux/fs.h. So we have to + * define the constants ourselves for now. + */ +#define FS_IOC_GETFLAGS _IOR('f', 1, long) +#define FS_IOC_SETFLAGS _IOW('f', 2, long) +#define FS_IOC_GETVERSION _IOR('v', 1, long) +#define FS_IOC_SETVERSION _IOW('v', 2, long) +#define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) +#define FS_IOC32_GETFLAGS _IOR('f', 1, int) +#define FS_IOC32_SETFLAGS _IOW('f', 2, int) +#define FS_IOC32_GETVERSION _IOR('v', 1, int) +#define FS_IOC32_SETVERSION _IOW('v', 2, int) +#else #include <linux/fs.h> +#endif #include <linux/fd.h> #if defined(CONFIG_FIEMAP) #include <linux/fiemap.h> diff --git a/meson.build b/meson.build index 294e9a8f32..30a380752c 100644 --- a/meson.build +++ b/meson.build @@ -1963,6 +1963,8 @@ config_host_data.set('HAVE_OPTRESET', cc.has_header_symbol('getopt.h', 'optreset')) config_host_data.set('HAVE_IPPROTO_MPTCP', cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP')) +config_host_data.set('HAVE_SYS_MOUNT_FSCONFIG', + cc.has_header_symbol('sys/mount.h', 'FSCONFIG_SET_FLAG')) # has_member config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID',
The latest glibc 2.36 has extended sys/mount.h so that it defines the FSCONFIG_* enum constants. These are historically defined in linux/mount.h, and thus if you include both headers the compiler complains: In file included from /usr/include/linux/fs.h:19, from ../linux-user/syscall.c:98: /usr/include/linux/mount.h:95:6: error: redeclaration of 'enum fsconfig_command' 95 | enum fsconfig_command { | ^~~~~~~~~~~~~~~~ In file included from ../linux-user/syscall.c:31: /usr/include/sys/mount.h:189:6: note: originally defined here 189 | enum fsconfig_command | ^~~~~~~~~~~~~~~~ /usr/include/linux/mount.h:96:9: error: redeclaration of enumerator 'FSCONFIG_SET_FLAG' 96 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ | ^~~~~~~~~~~~~~~~~ /usr/include/sys/mount.h:191:3: note: previous definition of 'FSCONFIG_SET_FLAG' with type 'enum fsconfig_command' 191 | FSCONFIG_SET_FLAG = 0, /* Set parameter, supplying no value */ | ^~~~~~~~~~~~~~~~~ ...snip... QEMU doesn't include linux/mount.h, but it does use linux/fs.h and thus gets linux/mount.h indirectly. glibc acknowledges this problem but does not appear to be intending to fix it in the forseeable future, simply documenting it as a known incompatibility with no workaround: https://sourceware.org/glibc/wiki/Release/2.36#Usage_of_.3Clinux.2Fmount.h.3E_and_.3Csys.2Fmount.h.3E https://sourceware.org/glibc/wiki/Synchronizing_Headers To address this requires either removing use of sys/mount.h or linux/fs.h, despite QEMU needing declarations from both. This patch removes linux/fs.h, meaning we have to define various FS_IOC constants that are now unavailable. Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> --- linux-user/syscall.c | 18 ++++++++++++++++++ meson.build | 2 ++ 2 files changed, 20 insertions(+)