efi: earlycon: Reprobe after parsing config tables
Ard Biesheuvel
maintainerPATCH 1/1 efi: earlycon: Reprobe after parsing config tables
Commit 732ea9db9d8a ("efi: libstub: Move screen_info handling to common code") reorganized the earlycon handling so that all architectures pass the screen_info data via a EFI config table instead of populating struct screen_info directly, as the latter is only possible when the EFI stub is baked into the kernel (and not into the decompressor).
However, this means that struct screen_info may not have been populated yet by the time the earlycon probe takes place, and this results in a non-functional early console.
So let's probe again right after parsing the config tables and populating struct screen_info. Note that this means that earlycon output starts a bit later than before, and so it may fail to capture issues that occur while doing the early EFI initialization.
--- drivers/firmware/efi/earlycon.c | 16 +++++++++++++--- drivers/firmware/efi/efi-init.c | 2 ++ include/linux/efi.h | 1 + 3 files changed, 16 insertions(+), 3 deletions(-)
drivers/firmware/efi/earlycon.c +13 -3
@@ -215,6 +215,14 @@ efi_earlycon_write(struct console *con, const char *str, unsigned int num) } } +static bool __initdata fb_probed; + +void __init efi_earlycon_reprobe (void ) +{ +if (fb_probed) +setup_earlycon ("efifb" ); +} +static int __init efi_earlycon_setup (struct earlycon_device *device,const char *opt) {
@@ -222,15 +230,17 @@ static int __init efi_earlycon_setup(struct earlycon_device *device,u16 xres, yres;u32 i; -if (screen_info.orig_video_isVGA !=VIDEO_TYPE_EFI ) + fb_wb = opt && !strcmp (opt,"ram" ); + +if (screen_info.orig_video_isVGA !=VIDEO_TYPE_EFI ) { + fb_probed = true;return -ENODEV ; + } fb_base = screen_info.lfb_base;if (screen_info.capabilities &VIDEO_CAPABILITY_64BIT_BASE ) fb_base |= (u64 )screen_info.ext_lfb_base <<32 ; - fb_wb = opt && !strcmp (opt,"ram" ); - si = &screen_info; xres = si->lfb_width; yres = si->lfb_height;
drivers/firmware/efi/efi-init.c +2 -0
@@ -72,6 +72,8 @@ static void __init init_screen_info(void)if (memblock_is_map_memory (screen_info.lfb_base))memblock_mark_nomap (screen_info.lfb_base, screen_info.lfb_size); + +efi_earlycon_reprobe (); } }
include/linux/efi.h +1 -0
@@ -693,6 +693,7 @@ efi_guid_to_str(efi_guid_t *guid, char *out) }extern void efi_init (void ); +extern void efi_earlycon_reprobe (void );#ifdef CONFIG_EFI extern void efi_enter_virtual_mode (void );/* switch EFI to virtual mode, if possible */ #else
Signature · 1 line
2.39.2
Quoted 16 lines · Ard Biesheuvel
Commit
732ea9db9d8a("efi: libstub: Movescreen_infohandling to common code") reorganized the earlycon handling so that all architectures pass thescreen_infodata via a EFI config table instead of populating structscreen_infodirectly, as the latter is only possible when the EFI stub is baked into the kernel (and not into the decompressor).However, this means that struct
screen_infomay not have been populated yet by the time the earlycon probe takes place, and this results in a non-functional early console.So let's probe again right after parsing the config tables and populating struct
screen_info. Note that this means that earlycon output starts a bit later than before, and so it may fail to capture issues that occur while doing the early EFI initialization.
Quoted 6 lines
Cc Xuefeng Li <[email protected]>Cc Xuerui Wang <[email protected]>Cc Huacai Chen <[email protected]>Reported-by Shawn Guo <[email protected]>Fixes 732ea9db9d8a ("efi: libstub: Move screen_info handling to common code")Signed-off-by Ard Biesheuvel <[email protected]>
Thanks for the fix! On Snapdragon laptops: