export ufs_fs.h to userspace
PATCH 1/1 export ufs_fs.h to userspace
was ufs_fs.h purposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship between ufs_fs.h and its sub headers (like ufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for example
---
include/linux/Kbuild +1 -0
@@ -324,6 +324,7 @@ unifdef-y += tty.h unifdef-y += types.h unifdef-y += udf_fs_i.h unifdef-y += udp.h +unifdef-y += ufs_fs.h unifdef-y += uinput.h unifdef-y += uio.h unifdef-y += unistd.h
include/linux/ufs_fs.h +3 -1
@@ -45,8 +45,10 @@ typedef __u32 __bitwise __fs32;typedef __u16 __bitwise __fs16 ;#endif +#ifdef __KERNEL__#include <linux/ufs_fs_i.h> #include <linux/ufs_fs_sb.h> +#endif #define UFS_BBLOCK 0#define UFS_BBSIZE 8192
@@ -303,7 +305,7 @@ typedef __u16 __bitwise __fs16;#define UFS_MAXMNTLEN 512#define UFS2_MAXMNTLEN 468#define UFS2_MAXVOLLEN 32 -/* #define UFS_MAXCSBUFS 31 */ +#define UFS_MAXCSBUFS 31#define UFS_LINK_MAX 32000/* #define UFS2_NOCSPTRS ((128 / sizeof(void *)) - 4)
include/linux/ufs_fs_sb.h +0 -1
@@ -21,7 +21,6 @@struct ufs_sb_private_info ;struct ufs_cg_private_info ;struct ufs_csum ; -#define UFS_MAXCSBUFS 31struct ufs_sb_info {struct ufs_sb_private_info * s_uspi;
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for example
are you sure it actually uses anything from this header, and not just assumes the magic number to be there??
(also.. I kind of would think it reasonable for things with their own UFS fs reader to have their own header)
Quoted 5 lines · Arjan van de Ven
Quoted 4 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes
Quoted 8 lines
advantage
Quoted 1 line
of this header for example
are you sure it actually uses anything from this header, and not just assumes the magic number to be there??
(also.. I kind of would think it reasonable for things with their own UFS fs reader to have their own header)
ufs_fs.h certainly has things which certain types of userspace programs could use:
on-disk layouts of UFS(2) inodes, cylinder groups, ... So it should be exported.
However, NAK patch in this form until all kernel internal things would dissapear from
exported header: struct ufs_buffer_head, ufs_sb_private_info. Probably something else, I haven't looked closely.
Quoted 11 lines · Arjan van de Ven
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for exampleare you sure it actually uses anything from this header, and not just assumes the magic number to be there??
(also.. I kind of would think it reasonable for things with their own UFS fs reader to have their own header)
silo utilizes the structures and the random defines so that it can query UFS filesystems directly
-mike
Quoted 8 lines · Mike Frysinger
Quoted 7 lines · Arjan van de Ven
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for example
Quoted 7 lines
Quoted 4 lines
are you sure it actually uses anything from this header, and not just assumes the magic number to be there??
(also.. I kind of would think it reasonable for things with their own UFS fs reader to have their own header)silo utilizes the structures and the random defines so that it can query UFS filesystems directly
at which point arguably it should have it's own UFS header. It's not using the UFS header to use as *kernel* interface!!
Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
Quoted 14 lines · Arjan van de Ven
Quoted 13 lines · Mike Frysinger
Quoted 11 lines · Arjan van de Ven
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for exampleare you sure it actually uses anything from this header, and not just assumes the magic number to be there??
(also.. I kind of would think it reasonable for things with their own UFS fs reader to have their own header)
Quoted 5 lines
Quoted 2 lines
silo utilizes the structures and the random defines so that it can query UFS filesystems directly
at which point arguably it should have it's own UFS header. It's not using the UFS header to use as *kernel* interface!!
true, and i wont argue that point too much other than to say it's nice that the kernel provides it so as to reduce code duplication in the world
in more realistic terms, not exporting the ufs_fs.h header prevents any sort of raw filesystem checkers which would be running in userspace, but then again i cant quote any such packages that actually exist ;)
in the end, it doesnt really matter to me right now one way or the other as the silo guys do include some ufs code to handle the case where it is being built by a non-linux toolchain
-mike
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for example
We should only export kernel interfaces and ufs_fs.h isn't one. silo wants it because it defines the ufs format - but the linux structs for that can and do change, e.g. adding unions when we add support for the gazillion+1st ufs format variant. silo should just grab a copy of it that it's happy with it and update it when it needs to support another format variant (which I think is a rather theoretical issue as silo only supports the solaris/sparc format IIR)
Quoted 7 lines · Christoph Hellwig
Quoted 5 lines · Mike Frysinger
was
ufs_fs.hpurposefully not exported to userspace or did it just slip through the cracks ? assuming the latter scenario, the attached patch touches up the relationship betweenufs_fs.hand its sub headers (likeufs_fs_sb.h) so that we can export it ... the silo bootloader takes advantage of this header for example
Quoted 7 lines
We should only export kernel interfaces and
ufs_fs.hisn't one. silo wants it because it defines the ufs format - but the linux structs for that can and do change, e.g. adding unions when we add support for the gazillion+1st ufs format variant. silo should just grab a copy of it that it's happy with it and update it when it needs to support another format variant (which I think is a rather theoretical issue as silo only supports the solaris/sparc format IIR)
seems logical to mean, thanks for the response
-mike
Quoted 9 lines · Mike Frysinger
Quoted 7 lines
We should only export kernel interfaces and
ufs_fs.hisn't one. silo wants it because it defines the ufs format - but the linux structs for that can and do change, e.g. adding unions when we add support for the gazillion+1st ufs format variant. silo should just grab a copy of it that it's happy with it and update it when it needs to support another format variant (which I think is a rather theoretical issue as silo only supports the solaris/sparc format IIR)seems logical to mean, thanks for the response
Btw, if you're interesting in helping out in this area a little bit, it would be nice if we could split up and move ufs_fs.h a little bit. We should have an:
fs/ufs/ufs.h for all driver internal declarations fs/ufs/ufs_layout.h that purely defines the layout can can be copied
directly into other tools that need the ufs layout like silo