[LEDE-DEV] [PATCH] bcm53xx: use iflag=skip_bytes for dd command during sysupgrade

Etienne Champetier champetier.etienne at gmail.com
Mon Dec 25 10:57:39 PST 2017


Hi Rafal,

2017-12-25 5:18 GMT-08:00 Rafał Miłecki <rafal at milecki.pl>:
> On 2017-12-24 00:26, Etienne Champetier wrote:
>>
>> 2017-12-23 14:14 GMT-08:00 Rafał Miłecki <zajec5 at gmail.com>:
>>>
>>> From: Rafał Miłecki <rafal at milecki.pl>
>>>
>>> Since BusyBox 1.25.0 dd command supports iflag=skip_bytes which allows
>>> skipping requested amount of bytes without reducing blocksize. Thanks to
>>> this we can leave default blocksize and let dd work more efficiently.
>>>
>>> On Netgear R6250 "dd skip=58 iflag=skip_bytes" can be 5 times faster
>>> than "dd bs=58 skip=1" when extracting TRX out of CHK.
>>>
>>> Signed-off-by: Rafał Miłecki <rafal at milecki.pl>
>>> ---
>>>  target/linux/bcm53xx/base-files/lib/upgrade/platform.sh | 4 ++--
>>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/target/linux/bcm53xx/base-files/lib/upgrade/platform.sh
>>> b/target/linux/bcm53xx/base-files/lib/upgrade/platform.sh
>>> index 1a33e3a447..06451f17fd 100644
>>> --- a/target/linux/bcm53xx/base-files/lib/upgrade/platform.sh
>>> +++ b/target/linux/bcm53xx/base-files/lib/upgrade/platform.sh
>>> @@ -274,11 +274,11 @@ platform_pre_upgrade() {
>>>  platform_trx_from_chk_cmd() {
>>>         local header_len=$((0x$(get_magic_long_at "$1" 4)))
>>>
>>> -       echo -n dd bs=$header_len skip=1
>>> +       echo -n dd skip=$header_len iflag=skip_bytes
>>>  }
>>>
>>>  platform_trx_from_cybertan_cmd() {
>>> -       echo -n dd bs=32 skip=1
>>> +       echo -n dd skip=32 iflag=skip_bytes
>>
>>
>> haven't checked, but would "tail -c+32" work (or maybe "tail -c+33") works
>> ?
>> I know I add a similar old dd version to handle in the past (but
>> nothing to do with OpenWRT) and "tail" was enough in the end
>
>
> I didn't test it, but I believe we could use tail somehow for that.
>
> The question is: why should we?
>
> We use "dd" in general for a lot of sysupgrade-related operations. It
> has more fearures like controlling a block size which we may want to use
> in a future for more optimizations. It can be used to "cut" a part of
> file from the middle of it.
>
> I don't see any real advantage of switching to tail.

Just checked and lede 17.01 also has busybox 1.25+, so no need for tail indeed



More information about the Lede-dev mailing list