[SLOF] [PATCH v3] boot: do not use catpad to concatenate strings

Alexey Kardashevskiy aik at ozlabs.ru
Mon Dec 11 19:02:18 AEDT 2017

On 11/12/17 17:30, Nikunj A Dadhania wrote:
> Hi Alexey,
> Alexey Kardashevskiy <aik at ozlabs.ru> writes:
>> On 08/12/17 16:32, Nikunj A Dadhania wrote:
>>> The catpad size is 1K size, which can be hit easily hit with around 20 devices
>>> with bootindex.
>>> Open code EVALUATE such that concatenation is not required. Replace usage of
>>> $cat with a dynamically allocated buffer(16K) here.
>>> Reported here: https://github.com/qemu/SLOF/issues/3
>>> Signed-off-by: Nikunj A Dadhania <nikunj at linux.vnet.ibm.com>
>> I've read the v1..v3 of this patch and (to my embarrassment) I do not
>> understand how this works at all :)
>> Above you said "concatenation is not required" but there is
>> bootdev-string-cat, how is that not concat? We did cat to bootdevice, now -
>> to bootdev-string-cat, but this is not just it?
> Earlier the concatenation was done using a common buffer $catpad
> slof/fs/base.fs:CREATE $catpad 400 allot
> Which had a limitation of 1K, my first patch was to increase the size of
> $catpad and Segher said that $catpad was kept small on purpose for
> concatenating quickly without much overhead.
> Version-2 patch did following things:
>     1) introduced a similar static buffer for bootdev concatenation without
>        using $cat.

This part I understood :)

>     2) Replaced concatenation in LOAD and LOAD-NEXT by open coding EVALUATE word.

This part I do not understand - why is this change needed or how does it
make it better?

There is a global list of boot devices - bootdevice; and for some reason
now there is another one - bootdev-buf, both are strings...

And there is also a load-list list which is what for? :)

> Version-3 made the static buffer allocate from heap instead of a static
> buffer.
>> For example:
>> 0 > $bootdev type  /pci at 800000020000000/ethernet at 1
>> /pci at 800000020000000/ethernet at 2 disk cdrom net net1  ok
> Regards
> Nikunj


More information about the SLOF mailing list