PB62x/740: root + sshd, ftpd, smbd, iptables, usbnet (needs testers)

[Markismus]

Lag
I have version 3 still installed and I disabled all features except ssh and the Inkpad 3 pro still felt laggy. So I disabled ssh, too and now it is snappy again. This does not have to be a causal relation, of course.

I do have rather large dictionaries installed, especially the Oxford 2nd edition is 200MB. So at first, I expected the lag to be dependent on this. However, the use of the dictionary itself is rather fast and the return of a lagless behaviour seems to argue against this.

Does anyone else experience an increase in lag? Did anyone experience it with older versions, but not with the newer one. Does anyone have other tweaks concerning this issue.

Wishlist: Minding the Store
What would the reasonably secure and least heavy approach be to get the store to show my home server over the internet?
Using LAN on my WiFi it is lightning fast: Effectively, tapping a book and rendering the download page is enough to time to download the epub. Even cbr's are downloaded within seconds. (This is the reason I am interested in VPN: Everything that works at home, works while using VPN, too.)
It uses only the native Pocketbook app, so there is no increase in footprint. That being said, the web browser ıtself is laggy. A better alternative would be nice.

[Marco77]

Hats off, SSH key-based auth works fine.

Last thing, why does PC-initiated rsync fail

sh: rsync: not found
rsync: connection unexpectedly closed (0 bytes received so far) [Receiver]
rsync error: remote command not found (code 127) at io.c(235) [Receiver=3.1.2]

, yet the binary is in the PATH?
(solved by specifying --rsync-path=/mnt/secure/bin/rsync)


@Markismus, check what process hogs CPU with htop (launch with TERM=xterm htop if it complains), sort by CPU% or look at the cumulative TIME.

Mine has eink-reader.app as top process, for instance
sftp-server: only if a long-standing SFTP transfer has occurred, which is a bit intensive for this SoC

[jelbo]

I just jailbreaked my PocketBook Touch HD 3 (PB632) and it worked great. I had to disable the firewall before I could reach it using SSH and SFTP etc. though.

[ezdiy]

Originally Posted by jelbo

I just jailbreaked my PocketBook Touch HD 3 (PB632) and it worked great. I had to disable the firewall before I could reach it using SSH and SFTP etc. though.

Very good! Was expecting far more trouble on smaller devices. As for the firewall, can you try adding the rules manually from ssh:

iptables -A INPUT -s 169.254.0.0/16 -i rndis0 -j ACCEPT
iptables -A INPUT -s 169.254.0.0/16 -i eth0 -j DROP

If that still cuts you off? If so, what does ifconfig looks like, presumably eth0/rndis0 is something else? The reason for using those rules is that 169.254 is "blessed" and doesn't ask for passwords, so it would be dangerous to expose it via wifi (but is fine via usbnet which is physically connected and never routed).

[jelbo]

It must have been another issue before because I just re-enabled the firewall, rebooted and can still access it using SSH and SFTP.

[ezdiy]

Originally Posted by Markismus

Lag
I have version 3 still installed and I disabled all features except ssh and the Inkpad 3 pro still felt laggy. So I disabled ssh, too and now it is snappy again. This does not have to be a causal relation, of course.

I just fail to reproduce this. One thing that does manifest as lag, however, is suspends. That white LED light on the bottom - it lits up when the cpu is spinning constantly, at least on 740-2 - I hope 740 does the same. When situation arises the cpu spins constantly (such as Wifi on), the device is less laggy to touchscreen taps. When it turns off after few secs after each touch, next touch is lagged - the device must wake up from suspend. This manifests as "sluggish" behavior.

The bottom line is that this lag is a good thing, it's absolutely crucial to battery life. With device spinned up all the time, battery dies much sooner.

[jelbo]

Speaking of power saving, I've enabled the option 'Don't suspend wifi when a service is in use'. After what amount of time would wifi normally disconnect? Right now, it seems to not disconnect at all for me, even when I'm connected on a public wifi spot and am sure nothing is connecting to my ereader from there.

It could be KOReader, maybe, even though that is not currently active.

[Markismus]

Originally Posted by ezdiy

One thing that does manifest as lag, however, is suspends. ...The bottom line is that this lag is a good thing, it's absolutely crucial to battery life. With device spinned up all the time, battery dies much sooner.

Supposing the lag is caused by suspend, I also find that changing the markers for highlighting and for page cropping can be extremely laggy. These are events that have a lot of prior touch events. Further the device sometimes fails to come out of suspend. (See this thread about quirky power behavior.)

I remember my desktop on Arch Linux having trouble pulling the then new AMD CPU Ryzen 2600 back to maximum performance after a suspend: basically remaining at 1.5GHz instead of 3.5GHz. I had to manually adjust it:

cpupower frequency-set --governor performance

. The trouble has disappeared with newer kernels.

What would be a good approach to test this on the Pocketbook 740-2? I am reading the Arch Linux Wiki, but can't find directly relevant information, yet. Did you already include tools in your package that could be used to troubleshoot this?

(My wife is already asking me when I am going to return the reader...)

[ezdiy]

Originally Posted by Markismus

Supposing the lag is caused by suspend, I also find that changing the markers for highlighting and for page cropping can be extremely laggy. These are events that have a lot of prior touch events. Further the device sometimes fails to come out of suspend. (See this thread about quirky power behavior.)

I remember my desktop on Arch Linux having trouble pulling the then new AMD CPU Ryzen 2600 back to maximum performance after a suspend: basically remaining at 1.5GHz instead of 3.5GHz. I had to manually adjust it:

cpupower frequency-set --governor performance

. The trouble has disappeared with newer kernels.

What would be a good approach to test this on the Pocketbook 740-2? I am reading the Arch Linux Wiki, but can't find directly relevant information, yet. Did you already include tools in your package that could be used to troubleshoot this?

(My wife is already asking me when I am going to return the reader...)

# echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

Try on pocketbook. The behavior you mention - setting stuff like backlight intensity is interesting, because it talks with monitor process (./pocketbook). This would imply that the command queue is clogged and monitor is slow to respond. A lot of stuff talks to this, and monitor also likes to snoop around the os and change behavior. Among other things, it checks if ssh is running.

[ezdiy]

Originally Posted by jelbo

Speaking of power saving, I've enabled the option 'Don't suspend wifi when a service is in use'. After what amount of time would wifi normally disconnect? Right now, it seems to not disconnect at all for me, even when I'm connected on a public wifi spot and am sure nothing is connecting to my ereader from there.

It could be KOReader, maybe, even though that is not currently active.

It's supposed to trigger only during lock, when screen is already off and showing zzzz. Further, browser must not be open, nor any other app that pings net (koreader might, it has notoriously bad pocketbook support for power handling).

One interesting bit about this is that nothing really changes when device screen is locked - except it now tries to turn off wifi now and then. When that happens, the setting will block the attempt if sessions are in progress. That's at least how its supposed to work, I tested it only few times, maybe something else breaks it now.

[jelbo]

Originally Posted by ezdiy

It's supposed to trigger only during lock, when screen is already off and showing zzzz. Further, browser must not be open, nor any other app that pings net (koreader might, it has notoriously bad pocketbook support for power handling).

One interesting bit about this is that nothing really changes when device screen is locked - except it now tries to turn off wifi now and then. When that happens, the setting will block the attempt if sessions are in progress. That's at least how its supposed to work, I tested it only few times, maybe something else breaks it now.

I have unrooted and reset the device to see what happens. It disconnects from the WiFi after a few seconds being idle on the homescreen. It did not disconnect when I was rooted, but maybe it was KOReader. I'll try rooting again in a few days.

[ezdiy]

Originally Posted by jelbo

I have unrooted and reset the device to see what happens. It disconnects from the WiFi after a few seconds being idle on the homescreen. It did not disconnect when I was rooted, but maybe it was KOReader. I'll try rooting again in a few days.

You don't need to remove it, just disable the settings option to revert back to wifi suspends on lock. Currently it's probably too aggresive in what it considers "service connected" (ie if you have some stale connections at the time of lock, it may never suspend wifi later). This is most likely because tcp timeouts don't really seem to work as they should, so I'll have to fix it with idle timeout/keepalvies or something.

[ormus23]

I've just tried this, with both v6 and v5 with a InkPad 3 (Model PB740), and Jailbreak isn't working.

"If it silently exits, chances are the exploit doesn't run on your firmware."

But, supposedly there has been success with the same model.

What can I do?

Thanks.

[ezdiy]

Originally Posted by ormus23

I've just tried this, with both v6 and v5 with a InkPad 3 (Model PB740), and Jailbreak isn't working.

"If it silently exits, chances are the exploit doesn't run on your firmware."

But, supposedly there has been success with the same model.

What can I do?

Thanks.

What's your firmware version?

[ormus23]

U740.5.17.1402

Haven't run any updates yet, so since you're bringing this up, I'm guessing that's what I must do?