Re: [oi-dev] libfuse update
Alexander Pyhalov wrote: On 08/26/16 12:52 PM, Alexander Pyhalov wrote: No surprise, as it's 755. On Linux I see 666. What is correct approach here? Should we somehow honor console user? Pardon, 600. And next thing which I receive is /usr/lib/fs/fusermount.bin: mount failed: Not owner I do not what this file is for, presumably for mounting as a plain user (which I never do). I have it as /usr/lib/fs/fuse/fusermount.bin with permissions 0555 owned by root:bin Jean-Pierre ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
Alexander Pyhalov wrote: On 08/26/16 12:27 PM, Jean-Pierre André wrote: Adam Števko wrote: Hi, [...] By the way, I sent several patches to libfuse to Adam last may. Did you take them into account ? Patches have been delivered since May: https://github.com/OpenIndiana/oi-userland/tree/oi/hipster/components/library/libfuse/patches Though, I have to go through our communication and check if that’s all. I recall I haven’t applied everything. I do not see the one I sent you on May 5th about permissions to set timestamps on files, and I do not remember anybody having objections about it (this has been in fuse-lite and used by ntfs-3g for years). I attach it again, just in case. BTW, when I run just-compiled /usr/lib/gvfs-fuse-daemon, I receive /dev/fuse: Permission denied. No surprise, as it's 755. On Linux I see 666. What is correct approach here? Should we somehow honor console user? /dev/fuse is the interface with the kernel module, created when the module is installed. I have always seen it with permissions 0666 to owner root:sys add_drv -m 'fuse 0666 root sys' fuse ln -s /devices/pseudo/fuse@0:fuse /dev/fuse (See http://jp-andre.pagesperso-orange.fr/openindiana-ntfs-3g.html) Jean-Pierre ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
On 08/26/16 12:52 PM, Alexander Pyhalov wrote: No surprise, as it's 755. On Linux I see 666. What is correct approach here? Should we somehow honor console user? Pardon, 600. And next thing which I receive is /usr/lib/fs/fusermount.bin: mount failed: Not owner -- Best regards, Alexander Pyhalov, system administrator of Southern Federal University IT department ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
On 08/26/16 12:51 PM, Alexander Pyhalov wrote: On 08/26/16 12:27 PM, Jean-Pierre André wrote: Adam Števko wrote: Hi, [...] By the way, I sent several patches to libfuse to Adam last may. Did you take them into account ? Patches have been delivered since May: https://github.com/OpenIndiana/oi-userland/tree/oi/hipster/components/library/libfuse/patches Though, I have to go through our communication and check if that’s all. I recall I haven’t applied everything. I do not see the one I sent you on May 5th about permissions to set timestamps on files, and I do not remember anybody having objections about it (this has been in fuse-lite and used by ntfs-3g for years). I attach it again, just in case. BTW, when I run just-compiled /usr/lib/gvfs-fuse-daemon, I receive /dev/fuse: Permission denied. No surprise, as it's 755. On Linux I see 666. What is correct approach here? Should we somehow honor console user? Pardon, 600. -- С уважением, Александр Пыхалов, программист отдела телекоммуникационной инфраструктуры управления информационно-коммуникационной инфраструктуры ЮФУ ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
On 08/26/16 12:27 PM, Jean-Pierre André wrote: Adam Števko wrote: Hi, [...] By the way, I sent several patches to libfuse to Adam last may. Did you take them into account ? Patches have been delivered since May: https://github.com/OpenIndiana/oi-userland/tree/oi/hipster/components/library/libfuse/patches Though, I have to go through our communication and check if that’s all. I recall I haven’t applied everything. I do not see the one I sent you on May 5th about permissions to set timestamps on files, and I do not remember anybody having objections about it (this has been in fuse-lite and used by ntfs-3g for years). I attach it again, just in case. BTW, when I run just-compiled /usr/lib/gvfs-fuse-daemon, I receive /dev/fuse: Permission denied. No surprise, as it's 755. On Linux I see 666. What is correct approach here? Should we somehow honor console user? -- Best regards, Alexander Pyhalov, system administrator of Southern Federal University IT department ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
Adam tevko wrote: Hi, [...] By the way, I sent several patches to libfuse to Adam last may. Did you take them into account ? Patches have been delivered since May: https://github.com/OpenIndiana/oi-userland/tree/oi/hipster/components/library/libfuse/patches Though, I have to go through our communication and check if thats all. I recall I havent applied everything. I do not see the one I sent you on May 5th about permissions to set timestamps on files, and I do not remember anybody having objections about it (this has been in fuse-lite and used by ntfs-3g for years). I attach it again, just in case. Jean-Pierre Adam ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev --- fuse.c.ref 2016-05-05 11:24:20.432471400 +0200 +++ fuse.c 2016-05-05 11:26:24.508680800 +0200 @@ -1462,6 +1462,29 @@ else err = fuse_fs_truncate(f->fs, path, attr->st_size); } +#ifdef HAVE_UTIMENSAT +if (!err && +(valid & (FUSE_SET_ATTR_ATIME | FUSE_SET_ATTR_MTIME))) { +struct timespec tv[2]; + +tv[0].tv_sec = 0; +tv[1].tv_sec = 0; +tv[0].tv_nsec = UTIME_OMIT; +tv[1].tv_nsec = UTIME_OMIT; + +if (valid & FUSE_SET_ATTR_ATIME_NOW) +tv[0].tv_nsec = UTIME_NOW; +else if (valid & FUSE_SET_ATTR_ATIME) +tv[0] = attr->st_atim; + +if (valid & FUSE_SET_ATTR_MTIME_NOW) +tv[1].tv_nsec = UTIME_NOW; +else if (valid & FUSE_SET_ATTR_MTIME) +tv[1] = attr->st_mtim; + +err = fuse_fs_utimens(f->fs, path, tv); +} else +#endif if (!err && (valid & (FUSE_SET_ATTR_ATIME | FUSE_SET_ATTR_MTIME)) == (FUSE_SET_ATTR_ATIME | FUSE_SET_ATTR_MTIME)) { struct timespec tv[2]; ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
Hi, > On Aug 26, 2016, at 10:20 AM, Jean-Pierre André > wrote: > > Alexander Pyhalov wrote: >> On 08/26/16 09:52 AM, Jean-Pierre André wrote: >> >>> What is the need ? AKAIK gvfs is related to Gnome and >>> is not supported on Linux any more. >> >> I see the following fuse-related errors, trying to compile gvfs with >> fuse support >> >> >> error: 'struct fuse_conn_info' has no member named 'want' > > This field exists in fuse 2.7, it is declared in the fuse-lite > library used by ntfs-3g. > >> error: 'FUSE_CAP_ATOMIC_O_TRUNC' undeclared > > This is one of the bits going into the field "want". Is is > about the file system telling it supports the O_TRUNC as an > argument of open()... which current fuse expects anyway. > > Both can be easily added to fuse_common.h, but this will > not make libfuse to 2.8. > > Patch proposal attached. > > By the way, I sent several patches to libfuse to Adam > last may. Did you take them into account ? Patches have been delivered since May: https://github.com/OpenIndiana/oi-userland/tree/oi/hipster/components/library/libfuse/patches Though, I have to go through our communication and check if that’s all. I recall I haven’t applied everything. Adam signature.asc Description: Message signed with OpenPGP using GPGMail ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
Alexander Pyhalov wrote: On 08/26/16 09:52 AM, Jean-Pierre André wrote: What is the need ? AKAIK gvfs is related to Gnome and is not supported on Linux any more. I see the following fuse-related errors, trying to compile gvfs with fuse support error: 'struct fuse_conn_info' has no member named 'want' This field exists in fuse 2.7, it is declared in the fuse-lite library used by ntfs-3g. error: 'FUSE_CAP_ATOMIC_O_TRUNC' undeclared This is one of the bits going into the field "want". Is is about the file system telling it supports the O_TRUNC as an argument of open()... which current fuse expects anyway. Both can be easily added to fuse_common.h, but this will not make libfuse to 2.8. Patch proposal attached. By the way, I sent several patches to libfuse to Adam last may. Did you take them into account ? Jean-Pierre As I see, gvfs is still supported: http://ftp.gnome.org/pub/GNOME/sources/gvfs/ --- fuse_common.h.old 2010-06-15 16:46:54.0 +0200 +++ fuse_common.h.new 2016-08-26 10:11:15.146085700 +0200 @@ -110,11 +110,23 @@ unsigned max_readahead; /** + * Capability flags, that the filesystem wants to enable + */ +unsigned int want; +/** * For future use. */ -unsigned reserved[27]; +unsigned reserved[26]; }; +/** + * Capability bits for 'fuse_conn_info.capable' and 'fuse_conn_info.want' + * + * FUSE_CAP_ATOMIC_O_TRUNC: filesystem handles the O_TRUNC open flag + */ +#define FUSE_CAP_ATOMIC_O_TRUNC (1 << 3) + + struct fuse_session; struct fuse_chan; ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
On 08/26/16 09:52 AM, Jean-Pierre André wrote: What is the need ? AKAIK gvfs is related to Gnome and is not supported on Linux any more. I see the following fuse-related errors, trying to compile gvfs with fuse support error: 'struct fuse_conn_info' has no member named 'want' error: 'FUSE_CAP_ATOMIC_O_TRUNC' undeclared As I see, gvfs is still supported: http://ftp.gnome.org/pub/GNOME/sources/gvfs/ -- Best regards, Alexander Pyhalov, system administrator of Southern Federal University IT department ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
Re: [oi-dev] libfuse update
Alexander Pyhalov wrote: Hi. Has someone looked at libfuse update? Even our ancient gvfs requires at least 2.8... Version 2.8 added ioctl() and poll(), both require significant updates to the fuse kernel module. IMHO ioctl() can be added with reasonable effort, it is available in the fuse-lite library shipped with ntfs-3g, but the kernel module part has to be created from scratch. poll() is more complex, I would not try unless there is a demand from someone who is blocked and has a real situation for testing. What is the need ? AKAIK gvfs is related to Gnome and is not supported on Linux any more. Jean-Pierre ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev
[oi-dev] libfuse update
Hi. Has someone looked at libfuse update? Even our ancient gvfs requires at least 2.8... -- System Administrator of Southern Federal University Computer Center ___ oi-dev mailing list oi-dev@openindiana.org https://openindiana.org/mailman/listinfo/oi-dev