@pjgowtham congratulations on finally having samba working
as for speeds... there could be a number of culprits here.
- your physical network speed.
i'm trying out enabling "hsdma" (DMA) and now iperf3 clocks at 933Mbit/sec (on a direct patch from the router to my laptop)... IIRC it was closer to 700Mbit/sec without hsdma. I'll be uploading that shortly. i'm not sure if it's just psychological, but i don't recall ever seeing > 300Mbit on 5G but i've seen 340Mbit rather frequently after enabling hsdma... 2.4G still peaks at 115 (sometimes 120) Mbit...
try downloading iperf3 on your pc (it's already pre-installed on my releases ;)) and see what you're getting (make sure that's not your bottleneck)
- USB device speed
granted i don't have a really clever way to test USB speed, on my latest bits i've done something really simple like plugging in a USB stick ("blue") and doing:
# time dd if=/dev/sda1 of=/dev/null bs=1M count=1000
1000+0 records in
1000+0 records out
real 0m 14.03s
user 0m 0.02s
sys 0m 8.77s
assuming i'm not making any false assumptions, that leads me to believe I can read 1000MB in 14 seconds which gives me a read speed of ~71MB/sec (i didn't try writing ;)) my
dmesg log says that write buffering is disabled for my device... what does yours say?
- Samba speed
I really have no idea what samba overhead is, but there must be some amount of overhead... plus you have ntfs-3g and fuse somewhere in between... those all add overhead. trying to ssh/copy a file I easily peg one of the CPUs on the router... run
top on your router while you're trying to copy files and see what's going on...
Having said all of the above, i would guess your bottleneck is 'samba' but test it...
EDIT: in case anyone wants to play around with slightly modified bits (i think i'm done with mt7615... i haven't seen a single panic and i think the speed is satisfactory... the only remaining problem is how you guys are going to get access to it once i push my r3p changes into stock openwrt.... i'm thinking about adding a package/module for mt7615 but that's probably going to be r3p-specific, which isn't a good idea in general.... but i don't have the ability or desire to test other platforms... or DBDC mode either...)
EDIT2: After seeing @lukasz92's post, I decided to compile the mt76/mt7615 module so we can have a baseline for comparing (ie, using the same openwrt/r3p bits we've been testing all along, and trying out the mt76-based module). It turns out it's much easier to compile than mediatek's closed-source code (suprise surprise!) and the kernel module packages are added to the "release" above. But beware, I have in no way tested them (and am too lazy to even load them now...)
EDIT3: @pellmen (always on the bleeding edge .. ;)) testing mt76 is now just an "opkg install" away...