Resetting grub_errno to GRUB_ERR_NONE before loading normal does allow
it to proceed.
At this point I'm not sure what is the correct behavior. Either
explicitly check for error and simply do not even try to automatically
jump into normal (it is still possible to load it manually) or reset
grub_errno and ignore.
The problem was exposed by someone trying to build minimal possible
image using grub-mkrescue and so omitting most partition modules. This
resulted in insmod for partition modules failure in embedded config.
I still tend to think the following patch matches current behavior - we
do not abort processing embedded config on error either. Alternative
would be to check and abort processing as soon as error is encountered.
From: Andrei Borzenkov <arvidj...@gmail.com>
Subject: [PATCH] kernel: reset grub_errno before loading normal
If last command in embedded config set grub_errno, loading of
normal fails. As we do not really check for any errors during
processing of embedded config and cannot handle error that happened
in the middle of it, just ignore error and hope for best.
---
grub-core/kern/main.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/grub-core/kern/main.c b/grub-core/kern/main.c
index 9cad0c4..2e46e2c 100644
--- a/grub-core/kern/main.c
+++ b/grub-core/kern/main.c
@@ -304,6 +304,8 @@ grub_main (void)
if (load_config)
grub_parser_execute (load_config);
+ /* Reset error, otherwise loading of normal fails */
+ grub_errno = GRUB_ERR_NONE;
grub_boot_time ("After execution of embedded config. Attempt to go to normal
mode");
--
tg: (018f79d..) u/reset-error-before-normal (depends on: master)
_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel
