[Samba] smbsrv_reply_printwrite not returning a response causing the print job to be truncated to 124 bytes

Thu Apr 29 10:42:50 UTC 2021

I haven't received last email but I can see your text appearing on the web
archive of the mailing list - I hope I'm replying in a correct way :)

I'm using the binaries from debian buster - the exact version
is 4.9.5+dfsg-5+deb10u1+rpi1 but I've also compiled latest stable release:
4.14.3. from what I could tell on github, last changes to this function
were made ~15 years ago

I've seen this issue has appeared already on the mailing list and I wrote
to the authors of the tickets and one of them responded that the last
version of samba that was working (that he checked) with this printing
protocol was 3.0.3. I have then looked inside the 3.0.3 source and the
function that implements this there is called reply_printwrite and at the
very beginning of the function I can see those two zeroes (one of them is
presumably 16 bit):

int outsize = set_message(outbuf,0,0,True);

which makes the response packet of 0xc1 + 3 zero bytes.

I'm starting samba via systemd:

sudo systemctl start smbd
sudo systemctl start nmbd

czw., 29 kwi 2021 o 08:45 Mateusz Mikołajczyk <mikolajczyk.mateusz at gmail.com>
napisał(a):

> yes, I'm aware that this protocol (and even this way of printing) is
> deprecated and believe me, I would like to move away from it :(
> unfortunetely that's the luxury I don't have.
>
> smb.conf is as following (lines proceeding with # removed for brevity,
> MINOLTA is the name of the CUPS printer (konica minolta)):
>
> ```
> [global]
>    workgroup = WORKGROUP
>    netbios name = SERVER
>    server string = SERVER
>    log file = /var/log/samba/log.%m
>    max log size = 1000
>    logging = file
>    panic action = /usr/share/samba/panic-action %d
>    server role = standalone server
>    obey pam restrictions = yes
>    unix password sync = yes
>    passwd program = /usr/bin/passwd %u
>    passwd chat = *Enter\snew\s*\spassword:* %n\n
> *Retype\snew\s*\spassword:* %n\n *password\supdated\ssuccessfully* .
>    pam password change = yes
>    map to guest = bad user
>
> [MINOLTA]
>    printer = MINOLTA
>    browseable = yes
>    printing = cups
>    path = /var/spool/samba
>    printable = yes
>    guest ok = yes
>    writable = yes
>    create mask = 0777
>
>    ntvfs handler = simple
>    aio read size = 0
>    aio write size = 0
>
>    strict sync = yes
>    sync always = yes
> ```
>
> p.s. it could be that this (i.e. synchronous reply with a 0xc1 packet and
> 3 null bytes) was an undocumented feature available only in the xp server,
> because when I tried to connect the DOS client to a windows 10 machine (of
> course I enabled smb1 / cifs protocol first) the print job would be added
> to the spooler but would indefinetely show 'buffering'.
>
>
> czw., 29 kwi 2021 o 08:35 Mateusz Mikołajczyk <
> mikolajczyk.mateusz at gmail.com> napisał(a):
>
>> yes, I'm aware that this protocol (and even this way of printing) is
>> deprecated and believe me, I would like to move away from it :(
>> unfortunetely that's the luxury I don't have.
>>
>> smb.conf is as following (lines proceeding with # removed for brevity,
>> MINOLTA is the name of the CUPS printer (konica minolta)):
>>
>> ```
>> [global]
>>    workgroup = WORKGROUP
>>    netbios name = SERVER
>>    server string = SERVER
>>    log file = /var/log/samba/log.%m
>>    max log size = 1000
>>    logging = file
>>    panic action = /usr/share/samba/panic-action %d
>>    server role = standalone server
>>    obey pam restrictions = yes
>>    unix password sync = yes
>>    passwd program = /usr/bin/passwd %u
>>    passwd chat = *Enter\snew\s*\spassword:* %n\n
>> *Retype\snew\s*\spassword:* %n\n *password\supdated\ssuccessfully* .
>>    pam password change = yes
>>    map to guest = bad user
>>
>> [MINOLTA]
>>    printer = MINOLTA
>>    browseable = yes
>>    printing = cups
>>    path = /var/spool/samba
>>    printable = yes
>>    guest ok = yes
>>    writable = yes
>>    create mask = 0777
>>
>>    ntvfs handler = simple
>>    aio read size = 0
>>    aio write size = 0
>>
>>    strict sync = yes
>>    sync always = yes
>> ```
>>
>> p.s. it could be that this (i.e. synchronous reply with a 0xc1 packet and
>> 3 null bytes) was an undocumented feature available only in the xp server,
>> because when I tried to connect the DOS client to a windows 10 machine (of
>> course I enabled smb1 / cifs protocol first) the print job would be added
>> to the spooler but would indefinetely show 'buffering'.
>>
>> czw., 29 kwi 2021 o 05:43 Jeremy Allison <jra at samba.org> napisał(a):
>>
>>> On Thu, Apr 29, 2021 at 01:38:04AM +0200, Mateusz Mikołajczyk via samba
>>> wrote:
>>> >Hello. I am investigating a weird issue with an old MS Net client for
>>> DOS
>>> >(v 3.0). I have used wireshark in order to inspect the packets and see
>>> the
>>> >difference between samba and windows xp server responses. this question
>>> is
>>> >only about printing capabilities as file sharing works flawlessly.
>>> >
>>> >when the client negotiates protocol it is settled as 2 / Greater than
>>> CORE
>>> >PROTOCOL and up to LANMAN2.1
>>> >
>>> >the start of print request seems to be OK, the server (in both
>>> >implementations) returns a file ID in the response that's labeled as
>>> 'Open
>>> >Print File Response / 0xC0'. Then the client sends print job (packet id:
>>> >0xc1) (124 bytes of it in fact - I suppose that's either due to some
>>> buffer
>>> >size or a maximum packet length) and that's when the servers differ in
>>> >terms of response.
>>> >
>>> >windows xp responds with 0xc1 and the data section contains 3 zero bytes
>>> >(0x00, 0x00, 0x00). Wireshark describes this as 'Write Print File
>>> >Response'. Samba does not respond at all which makes the client simply
>>> >hang. Then, after couple of seconds samba simply puts the print job to
>>> cups
>>> >which makes the output truncated. By inspecting samba source code for
>>> 0xc1
>>> >I saw a reference to smbsrv_reply_printwrite. I've read the source code
>>> for
>>> >that function and I can't understand why it would not return a response
>>> at
>>> >all. I would understand if it would return an error packet or something
>>> but
>>> >there's simply no response at all.
>>> >
>>> >here's what I tried so far:
>>> >
>>> >   ntvfs handler = print
>>> >   aio read size = 0
>>> >   aio write size = 0
>>> >
>>> >   strict sync = yes
>>> >   sync always = yes
>>> >
>>> >but somethinig tells me that because at the very beginning of the
>>> function
>>> >there's a hardcoded NTVFS_ASYNC_STATE_MAY_ASYNC then I am simply out of
>>> >luck?
>>>
>>> Looks like you're using the (unsupported code) ntvfs file
>>> server. That's legacy code only used for testing.
>>>
>>> You need to be talking to smbd, not the samba binary.
>>>
>>> Please ensure you have smbd running, and as I'm sure
>>> Rowland will ask, please submit your smb.conf :-).
>>>
>>
>>
>> --
>> pozdrawiam serdecznie,
>> Mateusz Mikołajczyk, a.k.a. toudi
>>
>
>
> --
> pozdrawiam serdecznie,
> Mateusz Mikołajczyk, a.k.a. toudi
>

-- 
pozdrawiam serdecznie,
Mateusz Mikołajczyk, a.k.a. toudi