Skip to contentSkip to conversation

Duplicate unicode tables in smb/cifs/jfs

DISCUSSION Dr. David Alan Gilbert 2 participants 4 messages Jun 25, 2023 – Jun 27, 2023 2 messages not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi All,
I just tripped over three sets of duplicated unicode tables and
wondered if anyone had tried to rationalise them:

The pair of:
./fs/smb/server/uniupr.h
./fs/smb/client/cifs_uniupr.h

are identical except for formatting.

./fs/jfs/jfs_uniupr.c,
and I think this is the same with some change in variable name.

(I'm guessing the same thing is implemented in a bunch of other places as well in a different way)

Would it make sense to swing fs/smb/server/uniupr.h up to hmm, maybe fs/uniupr.h, remove any of the cifs_ prefixes and then use the same include in all 3 places?

Maybe then later look at using some of the nls code?

Dave (who just tripped over this stuff)

Signature · 4 lines
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/
#1 lore raw
Quoted 4 lines · Dr. David Alan Gilbert

Hi All,
I just tripped over three sets of duplicated unicode tables and
wondered if anyone had tried to rationalise them:

Quoted 8 lines

The pair of:
./fs/smb/server/uniupr.h
./fs/smb/client/cifs_uniupr.h

are identical except for formatting.

./fs/jfs/jfs_uniupr.c,
and I think this is the same with some change in variable name.

From JFS's point of view, I wonder how relevant any of this code is. The Linux port of JFS originally was interested in compatibility with OS/2, which had case-insensitive file names. (Case-preserving, if I remember correctly, but names would match in a case-insensitive manner.)

I would be surprised if anyone cares about this feature anymore. Without a JFS_OS2 flag set in the superblock, none of the case-conversion code comes into play.

I assume SMB cares more about this and if Steve has an opinion on how to address this, I'd be happy to follow his lead. Probably better than ripping function out of JFS that could break some user that I'm not aware of.

Shaggy

Quoted 12 lines

(I'm guessing the same thing is implemented in a bunch of other places as well in a different way)

Would it make sense to swing fs/smb/server/uniupr.h up to hmm, maybe fs/uniupr.h, remove any of the cifs_ prefixes and then use the same include in all 3 places?

Maybe then later look at using some of the nls code?

Dave (who just tripped over this stuff)

A message that is not in the archive (sent off-list or not yet received): CAH2r5mvwZnd+S8CmZGQSdtnAWD45YjFx-1j0EaFBR3Qn-SjHEA@mail.gmail.com

#3 lore raw
Quoted 2 lines · * Steve French ([email protected]) wrote:

probably is low risk to make the ksmbd/cifs (server and client) code common for this

I'm up for trying to do that mod; do you have a feel of the best way? I guess this is two copies to avoid symbol clashes if both the server and clients are built/loaded on the same kernel?

I guess the clean way is to make this a separate .c/module that the others depend on and hten you have a nice single copy in the binary as well as the source.

The shorter hack that at least avoids the source dupe would be to change the declarations in the uniupr.h files to a #define that then instantiates it with different names in the place those are #included. You'd want to move the uniupr.h up somewhere, to hmm fs/uniupr.h ?

(Incidentally, I think the UNIUPR_NOLOWER is always defined so that whole chunk can go in both of them).

I guess the next level would be trying to merge parts of server/client unicode.[ch] - but I was just eyeing up this particularly simple dupe in that odd uniupr.h.

Quoted 1 line

(and leave the JFS code alone as Dave Kleikamp suggests)

Well hmm; my reading is the tables that JFS uses are *identical* to these; so if this header was somewhere outside of fs/smb it could probably #include it and just use the same table, with a no-binary-change.

Dave

Quoted 52 lines
Quoted 43 lines · Dave Kleikamp
Quoted 4 lines · Dr. David Alan Gilbert

Hi All,
I just tripped over three sets of duplicated unicode tables and
wondered if anyone had tried to rationalise them:

Quoted 8 lines

The pair of:
./fs/smb/server/uniupr.h
./fs/smb/client/cifs_uniupr.h

are identical except for formatting.

./fs/jfs/jfs_uniupr.c,
and I think this is the same with some change in variable name.

From JFS's point of view, I wonder how relevant any of this code is.
The Linux port of JFS originally was interested in compatibility with OS/2, which had case-insensitive file names. (Case-preserving, if I remember correctly, but names would match in a case-insensitive manner.)

I would be surprised if anyone cares about this feature anymore. Without a JFS_OS2 flag set in the superblock, none of the case-conversion code comes into play.

I assume SMB cares more about this and if Steve has an opinion on how to address this, I'd be happy to follow his lead. Probably better than ripping function out of JFS that could break some user that I'm not aware of.

Shaggy

Quoted 12 lines

(I'm guessing the same thing is implemented in a bunch of other places as well in a different way)

Would it make sense to swing fs/smb/server/uniupr.h up to hmm, maybe fs/uniupr.h, remove any of the cifs_ prefixes and then use the same include in all 3 places?

Maybe then later look at using some of the nls code?

Dave (who just tripped over this stuff)

Signature · 3 lines
Thanks,

Steve
Signature · 4 lines
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/

A message that is not in the archive (sent off-list or not yet received): CAH2r5mvEWeSOM=NYrxSG1EZ1py-DSkOrwEyh+_L-KuFLVWktQw@mail.gmail.com

#5 lore raw
Quoted 13 lines · * Steve French ([email protected]) wrote:
Quoted 9 lines · Dr. David Alan Gilbert
Quoted 1 line · * Steve French ([email protected]) wrote:

probably is low risk to make the ksmbd/cifs (server and client) code

common

Quoted 1 line

for this

I'm up for trying to do that mod; do you have a feel of the best way? I guess this is two copies to avoid symbol clashes if both the server and clients are built/loaded on the same kernel?

Quoted 7 lines

First step would be to move common Unicode/UCS-2 header definitions from fs/smb/client
and fs/smb/server to fs/smb/common

Second stuff could be having a common helper module in fs/smb/common similar to
fs/smb/common/cifs_md4.ko

OK, let me have a crack at that and see where I get to.

Dave

Quoted 95 lines
Quoted 89 lines

I guess the clean way is to make this a separate .c/module that the others depend on and hten you have a nice single copy in the binary as well as the source.

The shorter hack that at least avoids the source dupe would be to change the declarations in the uniupr.h files to a #define that then instantiates it with different names in the place those are #included. You'd want to move the uniupr.h up somewhere, to hmm fs/uniupr.h ?

(Incidentally, I think the UNIUPR_NOLOWER is always defined so that whole chunk can go in both of them).

I guess the next level would be trying to merge parts of server/client unicode.[ch] - but I was just eyeing up this particularly simple dupe in that odd uniupr.h.

Quoted 1 line

(and leave the JFS code alone as Dave Kleikamp suggests)

Well hmm; my reading is the tables that JFS uses are *identical* to these; so if this header was somewhere outside of fs/smb it could probably #include it and just use the same table, with a no-binary-change.

Dave

Quoted 22 lines

On Mon, Jun 26, 2023 at 12:03 PM Dave Kleikamp <[email protected]

Quoted 18 lines · wrote:
Quoted 4 lines · Dr. David Alan Gilbert

Hi All,
I just tripped over three sets of duplicated unicode tables and
wondered if anyone had tried to rationalise them:

Quoted 8 lines

The pair of:
./fs/smb/server/uniupr.h
./fs/smb/client/cifs_uniupr.h

are identical except for formatting.

./fs/jfs/jfs_uniupr.c,
and I think this is the same with some change in variable name.

From JFS's point of view, I wonder how relevant any of this code is.
The Linux port of JFS originally was interested in compatibility with OS/2, which had case-insensitive file names. (Case-preserving, if I remember correctly, but names would match in a case-insensitive

manner.)

Quoted 2 lines
Quoted 2 lines

I would be surprised if anyone cares about this feature anymore.

Without

Quoted 4 lines
Quoted 4 lines

a JFS_OS2 flag set in the superblock, none of the case-conversion code comes into play.

I assume SMB cares more about this and if Steve has an opinion on how

to

Quoted 21 lines
Quoted 4 lines

address this, I'd be happy to follow his lead. Probably better than ripping function out of JFS that could break some user that I'm not aware of.

Quoted 15 lines

Shaggy

Quoted 12 lines

(I'm guessing the same thing is implemented in a bunch of other places as well in a different way)

Would it make sense to swing fs/smb/server/uniupr.h up to hmm, maybe fs/uniupr.h, remove any of the cifs_ prefixes and then use the same include in all 3 places?

Maybe then later look at using some of the nls code?

Dave (who just tripped over this stuff)

Quoted 4 lines

--
Thanks,

Steve

--
-----Open up your eyes, open up your mind, open up your code -------

/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/
Signature · 3 lines
Thanks,

Steve
Signature · 4 lines
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/