Good day all,
Please join us today (Wednesday, March 27th) at 4PM EDT (20:00 UTC)
for the Fedora ARM weekly status meeting in #fedora-meeting-1 on Freenode.
On the agenda so far..
0) Status of ACTION items from our previous meeting
1) Problem packages
2) F19: uImage load addresses with unified kernel
3) Open Floor
If there is something that you would like to discuss that isn't mentioned
please feel free to bring it up at the end of the meeting or send an email
to the list.
Through Fedora 18 GA we've had a one-to-one mapping between the SoC and
the kernel rpm. Likewise, we produced a different filesystem image for
each SoC (EG, a highbank image, a panda image, a trimslice image, etc).
Since then the unified kernel in 3.7 and beyond has introduced the
ability to use a single kernel for multiple SoCs. This introduces the
pleasant prospect of having a single filesystem image that boots on
multiple ARM platforms. In fact, the standard unified kernel for F19
(3.9) may support highbank, omap, and versatile express all from the
same vmlinuz. Doing this still requires loading the right dbt, but I
have a plan for that. Doing this also requires creating a uImage with
the right load address, IE:
mkimage -A arm -O linux -T kernel -C none -a $ubootAddress
The trouble is, there is no unified $ubootAddress available. The
pandaboard uses 0x80008000, highbank and tegra use 0x00008000, a10 and
exynos5 use 0x40008000, and so forth. Not sure about beaglebone. If we
want to realize the dream of having a unified F19 release we need a
solution to this problem. I see three options:
1. Use the 0x00008000 address for highbank/tegra, demand bootz support
for everybody else. Since we control the panda/beagle/beaglebone uboots
they would be covered. Not clear on what this means for exynos5, but
exynos5 will use an LPAE kernel so that's a separate uimage already.
2. Generate multiple uImage files from a single kernel, then load the
appropriate uImage on boot.
3. Split up images to cope.
All 3 are viable, none are desirable. Is there a 4th option? If not,
what do want to go with?
Brendan Conoboy / Red Hat, Inc. / blc(a)redhat.com
Sorry if this is off-topic, but I have searched in Google and I cannot find an answer to this issue, or a hint to diagnose it further.
I have compiled a 3.5.6-1 kernel for a MCUZONE (www.mcuzone.com) board with a Fedora 17 armv5tel userspace. This particular board has two physical serial ports, of a maximum of three supported. The board boots with the kernel commandline:
console=ttyS2,115200n8 root=/dev/mmcblk0p2 rw rootwait ip=off . This results in systemd starting a serial console login with agetty in ttyS2, as normal and expected. For usability reasons, I want to run a second serial login in ttyS1. I therefore ran
"systemctl start 'serial-getty(a)ttyS1.service'", and the agetty starts. However, it will not complete the login process. It will show the banner and the login: prompt. I then type the user (say, "root"), and then it asks for the password, which I type
correctly, but then it complains "Inicio de sesión incorrecto" (Login incorrect). To rule out serial communication issues, I attached strace to each agetty in ttyS2 (working) and ttyS1 (not working), and both receive the username (one character at a time),
and then the password (as a single string followed by \n). So I am at a loss about why ttyS2 works and ttyS1 does not. As far as I can tell with strace, both instances receive the username and the password correctly. What can I do to diagnose this further?
Do I need to modify something else to enable the second login tty?