* [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel
@ 2024-06-20 7:16 Ahmad Fatoum
2024-06-20 7:39 ` Michael Taubert
2024-07-01 12:48 ` Sascha Hauer
0 siblings, 2 replies; 3+ messages in thread
From: Ahmad Fatoum @ 2024-06-20 7:16 UTC (permalink / raw)
To: barebox; +Cc: cybin, Ahmad Fatoum
Calling bootm on an x86 kernel image that lacks an EFI stub will result
in a "no image handler found for image type MBR sector", which can be
very confusing.
Improve upon this by directly suggesting that the kernel's
CONFIG_EFI_STUB option needs to be enabled, so the image has the
expected PE magic (and the EFI stub code that comes with it).
Cc: cybin <info@arachnodroid.de>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
---
efi/payload/image.c | 23 +++++++++++++++++++++++
1 file changed, 23 insertions(+)
diff --git a/efi/payload/image.c b/efi/payload/image.c
index 0795010cd9d2..6a41338f2a2f 100644
--- a/efi/payload/image.c
+++ b/efi/payload/image.c
@@ -337,6 +337,26 @@ static struct binfmt_hook binfmt_efi_hook = {
.hook = efi_execute,
};
+static int do_bootm_mbr(struct image_data *data)
+{
+ /* On x86, Linux kernel images have a MBR magic at the end of
+ * the first 512 byte sector and a PE magic if they're EFI-stubbed.
+ * The PE magic has precedence over the MBR, so if we arrive in
+ * this boot handler, the kernel has no EFI stub.
+ *
+ * Print a descriptive error message instead of "no image handler
+ * found for image type MBR sector".
+ */
+ pr_err("Can't boot MBR sector: Is CONFIG_EFI_STUB disabled in your Linux kernel config?\n");
+ return -ENOSYS;
+}
+
+static struct image_handler non_efi_handle_linux_x86 = {
+ .name = "non-EFI x86 Linux Image",
+ .bootm = do_bootm_mbr,
+ .filetype = filetype_mbr,
+};
+
static struct binfmt_hook binfmt_arm64_efi_hook = {
.type = filetype_arm64_efi_linux_image,
.hook = efi_execute,
@@ -350,6 +370,9 @@ static int efi_register_image_handler(void)
if (IS_ENABLED(CONFIG_CPU_V8))
binfmt_register(&binfmt_arm64_efi_hook);
+ if (IS_ENABLED(CONFIG_X86))
+ register_image_handler(&non_efi_handle_linux_x86);
+
return 0;
}
late_efi_initcall(efi_register_image_handler);
--
2.39.2
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel
2024-06-20 7:16 [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel Ahmad Fatoum
@ 2024-06-20 7:39 ` Michael Taubert
2024-07-01 12:48 ` Sascha Hauer
1 sibling, 0 replies; 3+ messages in thread
From: Michael Taubert @ 2024-06-20 7:39 UTC (permalink / raw)
To: Ahmad Fatoum, barebox
Tested-by: Michael Taubert <info@arachnodroid.de>
Am 20.06.24 um 09:16 schrieb Ahmad Fatoum:
> Calling bootm on an x86 kernel image that lacks an EFI stub will result
> in a "no image handler found for image type MBR sector", which can be
> very confusing.
>
> Improve upon this by directly suggesting that the kernel's
> CONFIG_EFI_STUB option needs to be enabled, so the image has the
> expected PE magic (and the EFI stub code that comes with it).
>
> Cc: cybin <info@arachnodroid.de>
> Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
> ---
> efi/payload/image.c | 23 +++++++++++++++++++++++
> 1 file changed, 23 insertions(+)
>
> diff --git a/efi/payload/image.c b/efi/payload/image.c
> index 0795010cd9d2..6a41338f2a2f 100644
> --- a/efi/payload/image.c
> +++ b/efi/payload/image.c
> @@ -337,6 +337,26 @@ static struct binfmt_hook binfmt_efi_hook = {
> .hook = efi_execute,
> };
>
> +static int do_bootm_mbr(struct image_data *data)
> +{
> + /* On x86, Linux kernel images have a MBR magic at the end of
> + * the first 512 byte sector and a PE magic if they're EFI-stubbed.
> + * The PE magic has precedence over the MBR, so if we arrive in
> + * this boot handler, the kernel has no EFI stub.
> + *
> + * Print a descriptive error message instead of "no image handler
> + * found for image type MBR sector".
> + */
> + pr_err("Can't boot MBR sector: Is CONFIG_EFI_STUB disabled in your Linux kernel config?\n");
> + return -ENOSYS;
> +}
> +
> +static struct image_handler non_efi_handle_linux_x86 = {
> + .name = "non-EFI x86 Linux Image",
> + .bootm = do_bootm_mbr,
> + .filetype = filetype_mbr,
> +};
> +
> static struct binfmt_hook binfmt_arm64_efi_hook = {
> .type = filetype_arm64_efi_linux_image,
> .hook = efi_execute,
> @@ -350,6 +370,9 @@ static int efi_register_image_handler(void)
> if (IS_ENABLED(CONFIG_CPU_V8))
> binfmt_register(&binfmt_arm64_efi_hook);
>
> + if (IS_ENABLED(CONFIG_X86))
> + register_image_handler(&non_efi_handle_linux_x86);
> +
> return 0;
> }
> late_efi_initcall(efi_register_image_handler);
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel
2024-06-20 7:16 [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel Ahmad Fatoum
2024-06-20 7:39 ` Michael Taubert
@ 2024-07-01 12:48 ` Sascha Hauer
1 sibling, 0 replies; 3+ messages in thread
From: Sascha Hauer @ 2024-07-01 12:48 UTC (permalink / raw)
To: barebox, Ahmad Fatoum; +Cc: cybin
On Thu, 20 Jun 2024 09:16:16 +0200, Ahmad Fatoum wrote:
> Calling bootm on an x86 kernel image that lacks an EFI stub will result
> in a "no image handler found for image type MBR sector", which can be
> very confusing.
>
> Improve upon this by directly suggesting that the kernel's
> CONFIG_EFI_STUB option needs to be enabled, so the image has the
> expected PE magic (and the EFI stub code that comes with it).
>
> [...]
Applied, thanks!
[1/1] efi: payload: inform users on attempt to boot non EFI-stubbed kernel
https://git.pengutronix.de/cgit/barebox/commit/?id=d9ec1ad67442 (link may not be stable)
Best regards,
--
Sascha Hauer <s.hauer@pengutronix.de>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-07-01 12:49 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-06-20 7:16 [PATCH] efi: payload: inform users on attempt to boot non EFI-stubbed kernel Ahmad Fatoum
2024-06-20 7:39 ` Michael Taubert
2024-07-01 12:48 ` Sascha Hauer
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox