Re: [PATCH 1/5] smb: common: split out smb_direct related header files

[Stefan Metzmacher <metze@samba.org>]

Am 27.05.25 um 20:50 schrieb Tom Talpey:
> I love the idea. Couple of questions on the pathnames...
> 
> On 5/27/2025 12:12 PM, Stefan Metzmacher wrote:
>> This is just a start moving into a common smb_direct layer.
>>
>> Cc: Steve French <smfrench@gmail.com>
>> Cc: Tom Talpey <tom@talpey.com>
>> Cc: Long Li <longli@microsoft.com>
>> Cc: Namjae Jeon <linkinjeon@kernel.org>
>> Cc: Hyunchul Lee <hyc.lee@gmail.com>
>> Cc: linux-cifs@vger.kernel.org
>> Cc: samba-technical@lists.samba.org
>> Signed-off-by: Stefan Metzmacher <metze@samba.org>
>> ---
>>   fs/smb/common/smb_direct/smb_direct.h     | 11 +++++
>>   fs/smb/common/smb_direct/smb_direct_pdu.h | 51 +++++++++++++++++++++++
> 
> Why the underscore in "smb_direct", in both components? The protocol
> doesn't use this, and it seems awkward and search-unfriendly.

Yes, I'd also prefer smbdirect and I just used it because I had
my existing wip driver under that name, but that should not matter.

The other reason was that the existing structures used smb_direct_
as prefix, but I'll also change that.

>>   fs/smb/server/transport_rdma.h            | 43 +------------------
>>   3 files changed, 64 insertions(+), 41 deletions(-)
>>   create mode 100644 fs/smb/common/smb_direct/smb_direct.h
>>   create mode 100644 fs/smb/common/smb_direct/smb_direct_pdu.h
>>
>> diff --git a/fs/smb/common/smb_direct/smb_direct.h b/fs/smb/common/smb_direct/smb_direct.h
>> new file mode 100644
>> index 000000000000..c745c37a3fea
>> --- /dev/null
>> +++ b/fs/smb/common/smb_direct/smb_direct.h
>> @@ -0,0 +1,11 @@
>> +/* SPDX-License-Identifier: GPL-2.0-or-later */
>> +/*
>> + *   Copyright (C) 2025, Stefan Metzmacher <metze@samba.org>
>> + */
>> +
>> +#ifndef __FS_SMB_COMMON_SMB_DIRECT_SMB_DIRECT_H__
>> +#define __FS_SMB_COMMON_SMB_DIRECT_SMB_DIRECT_H__
>> +
>> +#include "smb_direct_pdu.h"
> 
> And, why the empty redirection? It seems unnecessary, do I assume it
> will later contain API signatures for the planned common layer? Perhaps
> it should say this, to avoid confusion while that work is being done.

I'll think about it maybe I'll remove the redirection.

In the end the pdu definitions belong only in a single common .c files,
but it will take a lot of steps to get there.

I'll post a v2 tomorrow or next week.

Thanks!
metze