Menu
▾
▴
#2 GEF if bogus
Submitted by: Glen Foster
-------------------------------------------------------
CONFIGURATION:
The pertinent info is that I'm running recipe 5.5 on my
Eiger. Let me know if
you need other configuration information
-------------------------------------------------------
SYMPTOMS:
I accidentally fat-fingered the wrong "default" label
in elilo.conf, did a
quick write-and-quit-and-reboot on my Eiger. So elilo
couldn't find the correct
stanza (understandable since I didn't type the right
name). The result was a
General Exception Fault, which seems a bit extreme when
the EFI shel
l is also
avaiable to boot. I could understand the GEF if there
was nothing else for a
fall-back, but to have to reboot the box seems
overkill... besides, if the
machine booted the EFI shell, the administrator could
conceivably edit the
error in elilo.conf without having to go through the
rather leng
thy boot process.
What I got was:
Loading.: Red Hat Enterprise Linux AS
Starting: Red Hat Enterprise Linux AS
near line 26: default image 'linux-test' not defined
*** General Exception fault ***
Static General Registers
r000 0000000000000000 00000727ffb57000 r002
00000727ff894490 00000727ffb6d448
r004 0000000000000000 0000000000000000 r006
0000000000000001 0000000000000000
r008 00000727ffb644d8 00000040fefd9868 r010
00000040fefd9868 00000040fefef178
r012 00000727ffc308d0 0000000000000998 r014
00000040ff9f9c74 000000000000006c
r016 000000000000004c ffffffffffffffe0 r018
000000000000006c 0037003200320042
r020 00000040fef9c010 000000000000006c r022
0000000000000000 00000727ffb639a8
r024 00000000000000ff 0000000000001fff r026
0000000000002000 00000727ffb59228
r028 00000040ffe41fff 0000000000000007 r030
0000000000000008 0000000000000000
Stacked General Registers
r032 00000040fef9c010 0000000000000000 r034
0037003200320042 00000727ffc308d0
r036 00000040fefcd9c0 0000000000000288
Branch Registers
b000 00000040fefcd9c0 00000727ff39c820 b002
00000000ffc222b0 00000000ffc222b0
b004 0000000000000000 0000000000000000 b006
00000727ff894420 00000727ff894490
Kernel Registers
k000 0000000000000000 0000000000000000 k002
0000000000000000 0000000000000000
k004 0000000000000000 0000000000000000 k006
0000000000000000 0000000000000000
Misc Registers
itc = 0x0000002caec8ff35 itm =
0x0000002caf17c25e
iva = 0x00000727ff420000 isr =
0x0000000400000030
ifa = 0x0000000000000004 itir =
0x0000000000100030
iip = 0x00000727ff894550 iipa =
0x00000727ff894530
ip = 0x00000727ff894550 ipsr =
0x0000100001006038
bsp = 0x00000727ffc10620 bspstore =
0x00000727ffc104b0
rsc = 0x0000000000000000 pfs =
0x0000000000000288
cfm = 0x8000000000000306
cpu = 8 pr =
0x80000000ff6f9599
-------------------------------------------------------
CONDITIONS WHICH RELIABLY REPRODUCE BUG SYMPTOMS:
Setup a default boot entry (at least for Linux, I don't
know if it happens the
same with with other OSes) that doesn't have a
corresponding label= line in
elilo.conf and reboot your Eiger (running SFW recipe 5.5).
Logged In: YES
user_id=881514
elilo contained a bug where it was assuming at least one
parameter would be passed to it (the name of the elilo.efi
executable, usually). This assumption is true when elilo is
launched from the efi shell, or when it's launched from the
boot menu via netboot/tftp, but it is false when elilo is
launche d locally from the boot menu. That failed
assumption, along with the error in the config file, caused
elilo to call StriCmp with a bogus pointer.
The fix was simply to remove the "at least one argument"
assumption from the elilo.c (a single-line fix :o).
Fixed in cvs HEAD
(http://sourceforge.net/cvs/?group_id=91879), fix will be
included in elilo3.5pre2 release.
Logged In: YES
user_id=881514
elilo contained a bug where it was assuming at least one
parameter would be passed to it (the name of the elilo.efi
executable, usually). This assumption is true when elilo is
launched from the efi shell, or when it's launched from the
boot menu via netboot/tftp, but it is false when elilo is
launche d locally from the boot menu. That failed
assumption, along with the error in the config file, caused
elilo to call StriCmp with a bogus pointer.
The fix was simply to remove the "at least one argument"
assumption from the elilo.c (a single-line fix :o).
Fixed in cvs HEAD
(http://sourceforge.net/cvs/?group_id=91879), fix will be
included in elilo3.5pre2 release.