Forums

More fun in Linux La-La-Land.

Started by Unknown May 28, 2014
This is a follow-up on the attempt to fix a single key problem:

Problem:
Trying to build a USB bootable drive with Wifi and GSM. One of the Wifi driver is in the newer kernel only (3.6).  However, during boot, Linux complaints about "boot drive not ready", "press a key to wait, skip or manual" to continue.  The problem is in "mountall", but a "df" before "mountall" shows that the boot drive is alive and well.  So, the USB subsystem is not reporting something correctly to "mountall".

Solution:
It would have been easier to fix in older system, where "mountall" is simply a shell script.  However, not so simple in newer system. "mountall" is now a module with dbus, nih, etc. Just to get to compile the program, i have to work out:

python2.7-2.7.3 libpciaccess-0.13.2 dbus-1.6.18 libdrm-2.4.52 attr-2.4.47 gobject-introspection-1.40.0 gperf-3.0.4 intltool-0.50.2 libcap-2.24 libffi-3.0.13 libnih-1.0.3 plymouth-0.8.2 mountall-2.53

Nothing goes smoothly, some file in libdrm-2.4.52 is not compilable.  Still working on stubbing out the files.  And of course, installing dbus-1.6.18 broke the network manager, which remains to be fixed.

Question:
What module is handling the keyboard?  It might take me a few more days to find out, but someone might know it already.  People here could also be working on Embedded Linux.  Thanks.
On Wed, 28 May 2014, edward.ming.lee@gmail.com wrote:

> Date: Wed, 28 May 2014 07:04:12 -0700 (PDT) > From: edward.ming.lee@gmail.com > Newsgroups: comp.arch.embedded > Subject: More fun in Linux La-La-Land. > > This is a follow-up on the attempt to fix a single key problem: > > Problem: > Trying to build a USB bootable drive with Wifi and GSM. One of the Wifi driver is in the newer kernel only (3.6). However, during boot, Linux complaints about "boot drive not ready", "press a key to wait, skip or manual" to continue. The problem is in "mountall", but a "df" before "mountall" shows that the boot drive is alive and well. So, the USB subsystem is not reporting something correctly to "mountall". > > Solution: > It would have been easier to fix in older system, where "mountall" is simply a shell script. However, not so simple in newer system. "mountall" is now a module with dbus, nih, etc. Just to get to compile the program, i have to work out: > > python2.7-2.7.3 libpciaccess-0.13.2 dbus-1.6.18 libdrm-2.4.52 attr-2.4.47 gobject-introspection-1.40.0 gperf-3.0.4 intltool-0.50.2 libcap-2.24 libffi-3.0.13 libnih-1.0.3 plymouth-0.8.2 mountall-2.53 > > Nothing goes smoothly, some file in libdrm-2.4.52 is not compilable. Still working on stubbing out the files. And of course, installing dbus-1.6.18 broke the network manager, which remains to be fixed. > > Question: > What module is handling the keyboard? It might take me a few more days to find out, but someone might know it already. People here could also be working on Embedded Linux. Thanks. >
Edward, It is not entirely clear what you're attempting to accomplish, but it appears that you have not installed grub to the boot partition of the usb thumbdrive ... and possibly your boot partition is not created properly. There are a number of formulae for creating bootable drives on Linux, and my particular favourite includes using a virtualbox session running system rescue with gparted to partition the thumb drive in question, create the necessary file systems, format them, and then add grub from the command line to the boot partition of the thumb drive. I hope this helps ... Cheers, Rob.
On Wednesday, May 28, 2014 9:53:09 AM UTC-7, Sp...@ControlQ.com wrote:
> On Wed, 28 May 2014, edward.ming.lee@gmail.com wrote: > > > Date: Wed, 28 May 2014 07:04:12 -0700 (PDT) > > From: edward.ming.lee@gmail.com > > Newsgroups: comp.arch.embedded > > Subject: More fun in Linux La-La-Land. > > > This is a follow-up on the attempt to fix a single key problem: > > Problem: > > > Trying to build a USB bootable drive with Wifi and GSM. One of the Wifi driver is in the newer kernel only (3.6). However, during boot, Linux complaints about "boot drive not ready", "press a key to wait, skip or manual" to continue. The problem is in "mountall", but a "df" before "mountall" shows that the boot drive is alive and well. So, the USB subsystem is not reporting something correctly to "mountall". > > > Solution: > > > It would have been easier to fix in older system, where "mountall" is simply a shell script. However, not so simple in newer system. "mountall" is now a module with dbus, nih, etc. Just to get to compile the program, i have to work out: > > > python2.7-2.7.3 libpciaccess-0.13.2 dbus-1.6.18 libdrm-2.4.52 attr-2.4.47 gobject-introspection-1.40.0 gperf-3.0.4 intltool-0.50.2 libcap-2.24 libffi-3.0.13 libnih-1.0.3 plymouth-0.8.2 mountall-2.53 > > > Nothing goes smoothly, some file in libdrm-2.4.52 is not compilable. Still working on stubbing out the files. And of course, installing dbus-1.6.18 broke the network manager, which remains to be fixed. > > > Question: > > > What module is handling the keyboard? It might take me a few more days to find out, but someone might know it already. People here could also be working on Embedded Linux. Thanks. > > Edward, > > It is not entirely clear what you're attempting to accomplish, but it > appears that you have not installed grub to the boot partition of the usb > thumbdrive ... and possibly your boot partition is not created properly. > There are a number of formulae for creating bootable drives on Linux, and > my particular favourite includes using a virtualbox session running system > rescue with gparted to partition the thumb drive in question, create the > necessary file systems, format them, and then add grub from the command > line to the boot partition of the thumb drive. >
I am using lilo to boot. There is only one partition and it's marked active type 83. It boots and run properly, exact for asking to hit one key to continue. I can switch to grub if it will solve the problem, but i am not convinced that lilo is the problem.
On 28/05/14 19:01, edward.ming.lee@gmail.com wrote:
> On Wednesday, May 28, 2014 9:53:09 AM UTC-7, Sp...@ControlQ.com > wrote: >> On Wed, 28 May 2014, edward.ming.lee@gmail.com wrote: >> >>> Date: Wed, 28 May 2014 07:04:12 -0700 (PDT) From: >>> edward.ming.lee@gmail.com Newsgroups: comp.arch.embedded Subject: >>> More fun in Linux La-La-Land. >> >>> This is a follow-up on the attempt to fix a single key problem: >>> Problem: >> >>> Trying to build a USB bootable drive with Wifi and GSM. One of >>> the Wifi driver is in the newer kernel only (3.6). However, >>> during boot, Linux complaints about "boot drive not ready", >>> "press a key to wait, skip or manual" to continue. The problem >>> is in "mountall", but a "df" before "mountall" shows that the >>> boot drive is alive and well. So, the USB subsystem is not >>> reporting something correctly to "mountall". >> >>> Solution: >> >>> It would have been easier to fix in older system, where >>> "mountall" is simply a shell script. However, not so simple in >>> newer system. "mountall" is now a module with dbus, nih, etc. >>> Just to get to compile the program, i have to work out: >> >>> python2.7-2.7.3 libpciaccess-0.13.2 dbus-1.6.18 libdrm-2.4.52 >>> attr-2.4.47 gobject-introspection-1.40.0 gperf-3.0.4 >>> intltool-0.50.2 libcap-2.24 libffi-3.0.13 libnih-1.0.3 >>> plymouth-0.8.2 mountall-2.53 >> >>> Nothing goes smoothly, some file in libdrm-2.4.52 is not >>> compilable. Still working on stubbing out the files. And of >>> course, installing dbus-1.6.18 broke the network manager, which >>> remains to be fixed. >> >>> Question: >> >>> What module is handling the keyboard? It might take me a few >>> more days to find out, but someone might know it already. People >>> here could also be working on Embedded Linux. Thanks. >> >> Edward, >> >> It is not entirely clear what you're attempting to accomplish, but >> it appears that you have not installed grub to the boot partition >> of the usb thumbdrive ... and possibly your boot partition is not >> created properly. There are a number of formulae for creating >> bootable drives on Linux, and my particular favourite includes >> using a virtualbox session running system rescue with gparted to >> partition the thumb drive in question, create the necessary file >> systems, format them, and then add grub from the command line to >> the boot partition of the thumb drive. >> > > I am using lilo to boot. There is only one partition and it's marked > active type 83. It boots and run properly, exact for asking to hit > one key to continue. I can switch to grub if it will solve the > problem, but i am not convinced that lilo is the problem. >
I have no idea why you are having trouble doing something millions of other people have managed fine, why you think kernel 3.6 is "new", why you are posting to a Usenet group for embedded systems, or why you think LILO is a good choice of bootloader in the 21st century. I suspect you have a little knowledge of Linux from long ago, and are making things far more complicated for yourself as a result. Pretend you know nothing about Linux and consider yourself a complete newbie. As "the web" how you make a live bootable USB stick with a modern easy-to-use Linux distribution. My recommendation is Linux Mint - you can try the latest release candidate of version 17 from <http://www.linuxmint.com/>. The downloaded iso files can be put directly on a USB or a DVD - see <http://community.linuxmint.com/tutorial/view/744> for information.
> I have no idea why you are having trouble doing something millions of > other people have managed fine,
I don't know why you are always so negative.
> why you think kernel 3.6 is "new",
I don't know of any stable version newer than 3.6.18
> why you are posting to a Usenet group for embedded systems,
My target is an embedded system, probably the Intel Minnow Max board when it is ready.
> or why you think LILO is a good choice of bootloader in the 21st century.
LILO is simple and good enough for me. I don't need fancy boot options.
> I suspect you > have a little knowledge of Linux from long ago, and are making things > far more complicated for yourself as a result.
True.
> Pretend you know nothing about Linux and consider yourself a complete > newbie. As "the web" how you make a live bootable USB stick with a > modern easy-to-use Linux distribution. My recommendation is Linux Mint > - you can try the latest release candidate of version 17 from > <http://www.linuxmint.com/>. The downloaded iso files can be put > directly on a USB or a DVD - see
I have my base system from linuxmint version 15 and all the update modules are equivalent to version 17.
> > why you think kernel 3.6 is "new",
> I don't know of any stable version newer than 3.6.18
Sorry, i meant 3.13.6
> As "the web" how you make a live bootable USB stick with a > > modern easy-to-use Linux distribution.
I tried making USB key from LinuxMint 15 DVD before, it lockup during installation process. I don't have the external DVD drive to try newer version.
edward.ming.lee@gmail.com wrote:
> Trying to build a USB bootable drive with Wifi and GSM. One of the Wifi > driver is in the newer kernel only (3.6). However, during boot, Linux > complaints about "boot drive not ready", "press a key to wait, skip or > manual" to continue. The problem is in "mountall", but a "df" before > "mountall" shows that the boot drive is alive and well. So, the USB > subsystem is not reporting something correctly to "mountall".
If this is failing to mount the drive because it isn't alive yet (some USB sticks take a while to warm up) you can say rootdelay=30 on the kernel command line to give it 30s to come alive. rootwait will wait indefinitely. If it's further down the boot process (ie not the root device) you can add nobootwait as a mount option is /etc/fstab so it won't stall waiting for something to mount that isn't there. It could be you have a race: boot stalls because the root hasn't mounted, then the root mounts will sitting at the 'press to wait' prompt. One or other of these might help it. Theo
> If this is failing to mount the drive because it isn't alive yet (some USB > sticks take a while to warm up) you can say rootdelay=30 on the kernel > command line to give it 30s to come alive. rootwait will wait indefinitely.
I replace the "mountall" command with a script, calling: df mountall.exe (was mountall) "df" shows the root drive properly. The fact that this script is running shows that the drive is alive and well. But mountall.exe stills say it's not ready.
On 28/05/14 19:56, edward.ming.lee@gmail.com wrote:
>> I have no idea why you are having trouble doing something millions of >> other people have managed fine, > > I don't know why you are always so negative. >
I am not trying to be negative - sorry if I sound that way. (My recent posts in another couple of threads, from the same OP, have been worded rather negatively - they are based on substantial threads on other newsgroups, and I really do think that the OP there should be discouraged from his current plans.) If I give negative comments or criticism, it is with the best of intentions.
>> why you think kernel 3.6 is "new", > > I don't know of any stable version newer than 3.6.18
For many purposes, any recent kernel, including 3.6, is likely to be good enough. But the current release is 3.14.4, and I believe the latest with long-term support is 3.12.
> >> why you are posting to a Usenet group for embedded systems, > > My target is an embedded system, probably the Intel Minnow Max board when it is ready.
c.a.e. is mainly targeted at smaller embedded systems, with a discussion of hardware, software and other aspects of such systems. Embedded Linux systems are, obviously, "embedded" - but they are at the limit of the areas of interest for this group and don't come up often. For the most part, when you are using an off-the-shelf x86 based board for Linux, you will get best results by viewing it as a normal PC - particularly when you are talking about booting from a USB stick. There is also the group comp.os.linux.embedded, which might be a good place to look.
> >> or why you think LILO is a good choice of bootloader in the 21st century. > > LILO is simple and good enough for me. I don't need fancy boot options.
If you had got everything booting correctly, that might be fair enough. But LILO is an old bootloader - grub is much more popular and has a lot more features and flexibility. When you are trying to get things working on a newer card, it makes sense to use the tools everyone else uses. (On embedded Linux systems, especially non-x86 cards, the other popular choice is uboot.)
> >> I suspect you >> have a little knowledge of Linux from long ago, and are making things >> far more complicated for yourself as a result. > > True.
The trouble with this is that things change - and they can change quite rapidly in the Linux world. Techniques that worked well a few years ago can be outdated now - and sometimes you need to change to get things working with the latest hardware and the latest software. I know that I suffer from this - and I recognize it in others. There are times when the best solution is, as I said, to pretend you know nothing and do things the modern newbie way until you see how it fits together.
> >> Pretend you know nothing about Linux and consider yourself a complete >> newbie. As "the web" how you make a live bootable USB stick with a >> modern easy-to-use Linux distribution. My recommendation is Linux Mint >> - you can try the latest release candidate of version 17 from >> <http://www.linuxmint.com/>. The downloaded iso files can be put >> directly on a USB or a DVD - see > > I have my base system from linuxmint version 15 and all the update modules are equivalent to version 17. >
> For many purposes, any recent kernel, including 3.6, is likely to be > good enough. But the current release is 3.14.4, and I believe the > latest with long-term support is 3.12.
My mistake. I am using 3.13. Will try 3.14.
> >> why you are posting to a Usenet group for embedded systems,
> > My target is an embedded system, probably the Intel Minnow Max board when it is ready.
> c.a.e. is mainly targeted at smaller embedded systems, with a discussion > of hardware, software and other aspects of such systems. Embedded Linux > systems are, obviously, "embedded" - but they are at the limit of the > areas of interest for this group
I believe it's a hardware issue, the difference between SATA and USB MSD, and a software bug for reporting boot drive not ready. It works on SATA, but obviously does not work on USB. The desktop version Redhat, Ubuntu, Mint) are all geared for UI, which is not what i want for embedded board. The Dbus, NIH, plymouth, etc are tying everything together, including boot modules and network manager. It is a nightmare maintenance issue. I am trying to build a simpler system, at the expense of flexibility. I can fix/patch mountall by not checking the boot drive. It is ridiculous to report the drive containing the running program "not ready" anyway. However, the way "modern Linux" are setup. I have to compile the whole system to change one line. I will likely separate and/or remove the linkage between booting and other system managers.