Bad F17 install experiences: CentOS 6.3 vs. F17

[John Reiser]

> (1)
> The system has 3 NIC's, 2x Broadcom BCM5704 and 1x Netgear GA630.
> I use a BCM5704 for network install.  This works fine on CentOS with
> ksdevice=link.  With F17 I see multiple dhcp being started, resulting
> in multiple timeouts (I even see a second dhcp being started for em2
> while just before the console says that em1 got a link), and only
> after quite some time (a *long* time) the kickstart file is found

Please give some reasonable numerical estimate [say, +/- 20%]
of the duration: 30 seconds, 300 seconds (5 minutes), or what?

> and the install continues.

It is reasonable to start [attempt] a DHCP [in parallel] on each NIC.
It is better to assume that "ksdevice=link" indicates a preferred device,
and start that one first with shorter timeout (15 to 20 seconds),
perhaps even waiting for the preferred device before starting the others.
It is unreasonable not to proceed immediately with the first successful
reply.  Either kill the other DHCP, or let them proceed in parallel:
remember the results for use later if the first connection fails.

> 
> (2)
> The F17 installer can't start X and falls back to text mode, while
> the CentOS installer works fine in graphical mode.  The server has
> a simple onboard VGA controller.  This is what lspci says about the
> graphics controller:
> 
> VGA compatible controller: Advanced Micro Devices [AMD] nee ATI Rage XL (rev 27)
> 
> In the X.log I see:
> 
>     [   182.375] Backtrace:
>     [   182.375] 0: Xorg (xorg_backtrace+0x36) [0x464d46]
>     [   182.375] 1: Xorg (0x400000+0x69d99) [0x469d99]
>     [   182.375] 2: /lib64/libpthread.so.0 (0x7f315a2bf000+0xefe0) [0x7f315a2cdfe0]
>     [   182.375] 3: Xorg (UnloadSubModule+0xf) [0x475fbf]
>     [   182.375] 4: /usr/lib64/xorg/modules/drivers/mach64_drv.so (0x7f3157146000+0x183c3) [0x7f315715e3c3]
>     [   182.375] 5: Xorg (InitOutput+0x8ba) [0x4843ca]
>     [   182.375] 6: Xorg (0x400000+0x23246) [0x423246]
>     [   182.375] 7: /lib64/libc.so.6 (__libc_start_main+0xf5) [0x7f3158f46735]
>     [   182.375] 8: Xorg (0x400000+0x236bd) [0x4236bd]
>     [   182.375]
>     [   182.375] Segmentation fault at address (nil)
>     [   182.375]
>     Fatal server error:
>     [   182.375] Caught signal 11 (Segmentation fault). Server aborting

This is an ordinary bug, probably in some X11 driver, perhaps mach64_drv.
In practice for the last 2 years, I notice that X11 pays little attention
to anything except the "big 3": current Radeon, nVidia, and Intel graphics.
"3D" [alpha blending, etc.] has taken developer resources away from
any device which is 2D-only (cannot [does not] run Gnome shell.)

> [remainder snipped]

--