[SLOF] [PATCH v2 03/11] libnet: Pass tftp_retries and ip_version via struct filename_ip

Thomas Huth thuth at redhat.com
Fri May 25 15:34:01 AEST 2018


On 25.05.2018 07:11, Alexey Kardashevskiy wrote:
> On 25/5/18 3:08 pm, Thomas Huth wrote:
>> On 25.05.2018 07:01, Alexey Kardashevskiy wrote:
>>> On 25/5/18 2:28 pm, Thomas Huth wrote:
>>>> On 25.05.2018 04:57, Alexey Kardashevskiy wrote:
>>>>> On 25/5/18 12:26 am, Thomas Huth wrote:
>>>>>> On 24.05.2018 11:01, Alexey Kardashevskiy wrote:
>>>> [...]
>>>>> btw obp_tftp_args_t::filename[100] vs. filename_ip::filename[256] - it
>>>>> feels like the latter can be easily reduced to 140-ish (path + uuid) with
>>>>> no harm but probably this is for another day (both definitions came in the
>>>>> initial commit).
>>>>
>>>> filename_ip::filename can also come from the DHCP code, so decreasing
>>>
>>> Where else can it come from? And what can use it? TFTP can but what else?
>>
>> dhcpv4() in dhcp.c and dhcp6_process_options() in dhcpv6.c can set it,
>> too. tftp.c and pxelinux.c are the only consumers, as far as I know.
> 
> My point is if consumers cannot use all 256 bytes, then there is no point
> having this much in filename_ip. This is not to fix now though, I am just
> commenting.

As far as I can see, the boot file name in dhcpv4 can be up to 255
bytes, and the one from dhcpv6 could even be longer, so 256 sounds like
an adequate size for filename to me.

 Thomas


More information about the SLOF mailing list