Skip to contentSkip to conversation

efi/libstub: fix efi_parse_options() ignoring the default command line

PATCH Superseded Jonathan Marek 3 participants 15 messages Oct 11, 2024 – Oct 15, 2024 EFI

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 efi/libstub: fix efi_parse_options() ignoring the default command line

efi_convert_cmdline() always returns a size of at least 1 because it counts the NUL terminator, so the "cmdline_size == 0" condition is not possible.

Change it to compare against 1 to get the intended behavior: to use CONFIG_CMDLINE when load_options_size is 0.

Fixes 60f38de7a8d4 ("efi/libstub: Unify command line param parsing")
---
 drivers/firmware/efi/libstub/efi-stub.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/firmware/efi/libstub/efi-stub.c +1 -1
index 958a680e0660d..709ae2d41a632 100644
@@ -129,7 +129,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 
 	if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
 	    IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
-	    cmdline_size == 0) {
+	    cmdline_size == 1) {
 		status = efi_parse_options(CONFIG_CMDLINE);
 		if (status != EFI_SUCCESS) {
 			efi_err("Failed to parse options\n");
Signature · 1 line
2.45.1
#1 lore raw

PATCH 2/3 efi/libstub: remove uneccessary cmdline_size init/check

efi_convert_cmdline() always sets cmdline_size to at least 1 on success, so both the initialization to 0 and > 0 comparison are unecessary.

---
 drivers/firmware/efi/libstub/efi-stub.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/firmware/efi/libstub/efi-stub.c +2 -2
index 709ae2d41a632..f166614ef8432 100644
@@ -112,7 +112,7 @@ static u32 get_supported_rt_services(void)
 
 efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 {
-	int cmdline_size = 0;
+	int cmdline_size;
 	efi_status_t status;
 	char *cmdline;
 
@@ -137,7 +137,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 		}
 	}
 
-	if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_size > 0) {
+	if (!IS_ENABLED(CONFIG_CMDLINE_FORCE)) {
 		status = efi_parse_options(cmdline);
 		if (status != EFI_SUCCESS) {
 			efi_err("Failed to parse options\n");
Signature · 1 line
2.45.1
maintainer
#2 lore raw

Re: [PATCH 2/3] efi/libstub: remove uneccessary cmdline_size init/check

Quoted 4 lines · Jonathan Marek

efi_convert_cmdline() always sets cmdline_size to at least 1 on success, so both the initialization to 0 and > 0 comparison are unecessary.

The intent is to avoid parsing the empty string, so arguably, we should test for cmdline_size > 1 instead. But if we fix efi_convert_cmdline() instead, as I proposed in reply to 1/3, I guess we can drop this patch.

Quoted 30 lines
---
 drivers/firmware/efi/libstub/efi-stub.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/firmware/efi/libstub/efi-stub.c +2 -3
index 709ae2d41a632..f166614ef8432 100644
@@ -112,7 +112,7 @@ static u32 get_supported_rt_services(void)
 
 efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 {
-       int cmdline_size = 0;
+       int cmdline_size;
        efi_status_t status;
        char *cmdline;
 
@@ -137,7 +137,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
                }
        }
 
-       if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_size > 0) {
+       if (!IS_ENABLED(CONFIG_CMDLINE_FORCE)) {
                status = efi_parse_options(cmdline);
                if (status != EFI_SUCCESS) {
                        efi_err("Failed to parse options\n");
--

2.45.1

#3 lore raw

PATCH 3/3 efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

In reply to Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

---
 drivers/firmware/efi/libstub/file.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/firmware/efi/libstub/file.c +12 -0
index d6a025df07dcf..2a69e2b3583d4 100644
@@ -208,6 +208,18 @@ efi_status_t handle_cmdline_files(efi_loaded_image_t *image,
 	if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
 		efi_chunk_size = EFI_READ_CHUNK_SIZE;
 
+	if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
+	    IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
+	    cmdline_len == 0) {
+		if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_len > 0) {
+			/* both CONFIG_CMDLINE and load_options should be used */
+			efi_warn("ignoring %ls from CONFIG_CMDLINE\n", optstr);
+		} else {
+			cmdline = L"" CONFIG_CMDLINE;
+			cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
+		}
+	}
+
 	alloc_addr = alloc_size = 0;
 	do {
 		struct finfo fi;
Signature · 1 line
2.45.1
maintainer
#4 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

Quoted 8 lines · Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

The core kernel has its own handling for EXTEND/FORCE, so while we should parse it in the EFI stub to look for options that affect the stub's own behavior, we should not copy it into the command line that the stub provides to the core kernel.

E.g., drivers/of/fdt.c takes the bootargs from the DT and combines them with CONFIG_CMDLINE.

Quoted 31 lines
---
 drivers/firmware/efi/libstub/file.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/firmware/efi/libstub/file.c +12 -1
index d6a025df07dcf..2a69e2b3583d4 100644
@@ -208,6 +208,18 @@ efi_status_t handle_cmdline_files(efi_loaded_image_t *image,
        if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
                efi_chunk_size = EFI_READ_CHUNK_SIZE;
 
+       if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
+           IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
+           cmdline_len == 0) {
+               if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_len > 0) {
+                       /* both CONFIG_CMDLINE and load_options should be used */
+                       efi_warn("ignoring %ls from CONFIG_CMDLINE\n", optstr);
+               } else {
+                       cmdline = L"" CONFIG_CMDLINE;
+                       cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
+               }
+       }
+
        alloc_addr = alloc_size = 0;
        do {
                struct finfo fi;
--

2.45.1

#5 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

Quoted 10 lines · Ard Biesheuvel
Quoted 8 lines · Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

Quoted 9 lines

The core kernel has its own handling for EXTEND/FORCE, so while we should parse it in the EFI stub to look for options that affect the stub's own behavior, we should not copy it into the command line that the stub provides to the core kernel.

E.g., drivers/of/fdt.c takes the bootargs from the DT and combines them with CONFIG_CMDLINE.

I'm aware of that - the replacement the commit message is referring to is specifically for handle_cmdline_files() which this commit is modifying.

Currently efistub completely ignores initrd= and dtb= options provided through CONFIG_CMDLINE (handle_cmdline_files() only parses the EFI options)

For the EXTEND case, I didn't implement the full solution because its more complex and EXTEND is not available on arm64 anyway, so I went with just printing a warning instead.

maintainer
#6 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

Quoted 19 lines · Jonathan Marek
Quoted 15 lines · Ard Biesheuvel
Quoted 8 lines · Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

The core kernel has its own handling for EXTEND/FORCE, so while we should parse it in the EFI stub to look for options that affect the stub's own behavior, we should not copy it into the command line that the stub provides to the core kernel.

Quoted 8 lines
Quoted 4 lines

E.g., drivers/of/fdt.c takes the bootargs from the DT and combines them with CONFIG_CMDLINE.

I'm aware of that - the replacement the commit message is referring to is specifically for handle_cmdline_files() which this commit is modifying.

Ah ok - I missed that.

This is the kind of context that I'd expect in a cover letter, i.e., that the command line handling is inconsistent, and that we obtain the command line from the loaded image twice.

Also, the fact the initrd= handling and dtb= are special, because
a) multiple initrd= arguments are processed in order, and the files concatenated,
b) the filenames are consumed as UTF-16 as they are plugged into the file I/O protocols

Quoted 3 lines

Currently efistub completely ignores initrd= and dtb= options provided through CONFIG_CMDLINE (handle_cmdline_files() only parses the EFI options)

Indeed. You haven't explained why this is a problem. initrd= and dtb= contain references to files in the file system, and this does not seem like something CONFIG_EXTEND was intended for.

Quoted 3 lines

For the EXTEND case, I didn't implement the full solution because its more complex and EXTEND is not available on arm64 anyway, so I went with just printing a warning instead.

This code is shared between all architectures, so what arm64 does or does not support is irrelevant.

Can you explain your use case please?

#7 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

Quoted 35 lines · Ard Biesheuvel
Quoted 19 lines · Jonathan Marek
Quoted 15 lines · Ard Biesheuvel
Quoted 8 lines · Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

The core kernel has its own handling for EXTEND/FORCE, so while we should parse it in the EFI stub to look for options that affect the stub's own behavior, we should not copy it into the command line that the stub provides to the core kernel.

Quoted 8 lines
Quoted 4 lines

E.g., drivers/of/fdt.c takes the bootargs from the DT and combines them with CONFIG_CMDLINE.

I'm aware of that - the replacement the commit message is referring to is specifically for handle_cmdline_files() which this commit is modifying.

Ah ok - I missed that.

This is the kind of context that I'd expect in a cover letter, i.e., that the command line handling is inconsistent, and that we obtain the command line from the loaded image twice.

Quoted 6 lines

Also, the fact the initrd= handling and dtb= are special, because
a) multiple initrd= arguments are processed in order, and the files concatenated,
b) the filenames are consumed as UTF-16 as they are plugged into the file I/O protocols

(not relevant to this commit, but I need to say that concatenating dtb files makes no sense, only the first one will be used by the kernel)

Quoted 8 lines
Quoted 3 lines

Currently efistub completely ignores initrd= and dtb= options provided through CONFIG_CMDLINE (handle_cmdline_files() only parses the EFI options)

Indeed. You haven't explained why this is a problem. initrd= and dtb= contain references to files in the file system, and this does not seem like something CONFIG_EXTEND was intended for.

Its not the expected/documented behavior, that should be enough to make it a problem. Nowhere is it documented that these options would be ignored if provided through CONFIG_CMDLINE.

Quoted 9 lines
Quoted 3 lines

For the EXTEND case, I didn't implement the full solution because its more complex and EXTEND is not available on arm64 anyway, so I went with just printing a warning instead.

This code is shared between all architectures, so what arm64 does or does not support is irrelevant.

Can you explain your use case please?

I boot linux as the "EFI/Boot/bootaa64.efi" on my EFI partition. The firmware does not provide any load options. This system needs a dtb, so I add the dtb to my EFI partition and configure it using the dtb= option (using CONFIG_CMDLINE).

maintainer
#8 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

Quoted 44 lines · Jonathan Marek
Quoted 35 lines · Ard Biesheuvel
Quoted 19 lines · Jonathan Marek
Quoted 15 lines · Ard Biesheuvel
Quoted 8 lines · Jonathan Marek

Replace cmdline with CONFIG_CMDLINE when it should be used instead of load_options.

In the EXTEND case, it may be necessary to combine both CONFIG_CMDLINE and load_options. In that case, keep the old behavior and print a warning about the incorrect behavior.

The core kernel has its own handling for EXTEND/FORCE, so while we should parse it in the EFI stub to look for options that affect the stub's own behavior, we should not copy it into the command line that the stub provides to the core kernel.

Quoted 8 lines
Quoted 4 lines

E.g., drivers/of/fdt.c takes the bootargs from the DT and combines them with CONFIG_CMDLINE.

I'm aware of that - the replacement the commit message is referring to is specifically for handle_cmdline_files() which this commit is modifying.

Ah ok - I missed that.

This is the kind of context that I'd expect in a cover letter, i.e., that the command line handling is inconsistent, and that we obtain the command line from the loaded image twice.

Quoted 6 lines

Also, the fact the initrd= handling and dtb= are special, because
a) multiple initrd= arguments are processed in order, and the files concatenated,
b) the filenames are consumed as UTF-16 as they are plugged into the file I/O protocols

Quoted 3 lines

(not relevant to this commit, but I need to say that concatenating dtb files makes no sense, only the first one will be used by the kernel)

Sure, but this code was written for initrd= initially, and was reused for dtb=

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

Currently efistub completely ignores initrd= and dtb= options provided through CONFIG_CMDLINE (handle_cmdline_files() only parses the EFI options)

Quoted 9 lines
Quoted 4 lines

Indeed. You haven't explained why this is a problem. initrd= and dtb= contain references to files in the file system, and this does not seem like something CONFIG_EXTEND was intended for.

Its not the expected/documented behavior, that should be enough to make it a problem. Nowhere is it documented that these options would be ignored if provided through CONFIG_CMDLINE.

Fair enough.

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

For the EXTEND case, I didn't implement the full solution because its more complex and EXTEND is not available on arm64 anyway, so I went with just printing a warning instead.

This code is shared between all architectures, so what arm64 does or does not support is irrelevant.

Quoted 7 lines
Quoted 2 lines

Can you explain your use case please?

I boot linux as the "EFI/Boot/bootaa64.efi" on my EFI partition. The firmware does not provide any load options. This system needs a dtb, so I add the dtb to my EFI partition and configure it using the dtb= option (using CONFIG_CMDLINE).

Right.

Would the below work for you? It's not the prettiest code in the world, but at least it keeps the weird local to the function.

drivers/firmware/efi/libstub/file.c +10 -5
@@ -189,26 +189,48 @@
                                  unsigned long *load_addr,
                                  unsigned long *load_size)
 {
-       const efi_char16_t *cmdline = efi_table_attr(image, load_options);
-       u32 cmdline_len = efi_table_attr(image, load_options_size);
        unsigned long efi_chunk_size = ULONG_MAX;
        efi_file_protocol_t *volume = NULL;
+       const efi_char16_t *cmdline;
        efi_file_protocol_t *file;
        unsigned long alloc_addr;
        unsigned long alloc_size;
        efi_status_t status;
+       bool again = false;
+       u32 cmdline_len;
        int offset;
 
        if (!load_addr || !load_size)
                return EFI_INVALID_PARAMETER;
 
-       efi_apply_loadoptions_quirk((const void **)&cmdline, &cmdline_len);
-       cmdline_len /= sizeof(*cmdline);
-
        if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
                efi_chunk_size = EFI_READ_CHUNK_SIZE;
 
        alloc_addr = alloc_size = 0;
+
+#ifdef CONFIG_CMDLINE
+       if (IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
+           IS_ENABLED(CONFIG_CMDLINE_OVERRIDE) ||
+           (again = (IS_ENABLED(CONFIG_X86) ||
+                     IS_ENABLED(CONFIG_CMDLINE_EXTEND)))) {
+               static const efi_char16_t builtin_cmdline[] = L""

CONFIG_CMDLINE;
+
+ cmdline = builtin_cmdline;
+ cmdline_len = sizeof(builtin_cmdline);
+ } else
+#endif
+ {
+do_load_options:
+ cmdline = efi_table_attr(image, load_options);
+ cmdline_len = efi_table_attr(image, load_options_size);
+
+ efi_apply_loadoptions_quirk((const void **)&cmdline,
+ &cmdline_len);
+
+ again = false;
+ }
+ cmdline_len /= sizeof(*cmdline);
+

        do {
                struct finfo fi;
                unsigned long size;
@@ -290,6 +312,9 @@
                efi_call_proto(volume, close);
        } while (offset > 0);

+ if (again)
+ goto do_load_options;
+

        *load_addr = alloc_addr;
        *load_size = alloc_size;
#9 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

On 10/12/24 10:50 AM, Ard Biesheuvel wrote:
...

Quoted 6 lines

Right.

Would the below work for you? It's not the prettiest code in the world, but at least it keeps the weird local to the function.

Its missing the "load_options_size == 0" case, then CONFIG_CMDLINE should be used even if FORCE/etc. are not set. Otherwise it looks OK. (cmdline_len also shouldn't include the NUL character, but I don't think that matters)

Quoted 67 lines
drivers/firmware/efi/libstub/file.c +27 -5
@@ -189,26 +189,48 @@
                                   unsigned long *load_addr,
                                   unsigned long *load_size)
  {
-       const efi_char16_t *cmdline = efi_table_attr(image, load_options);
-       u32 cmdline_len = efi_table_attr(image, load_options_size);
         unsigned long efi_chunk_size = ULONG_MAX;
         efi_file_protocol_t *volume = NULL;
+       const efi_char16_t *cmdline;
         efi_file_protocol_t *file;
         unsigned long alloc_addr;
         unsigned long alloc_size;
         efi_status_t status;
+       bool again = false;
+       u32 cmdline_len;
         int offset;
 
         if (!load_addr || !load_size)
                 return EFI_INVALID_PARAMETER;
 
-       efi_apply_loadoptions_quirk((const void **)&cmdline, &cmdline_len);
-       cmdline_len /= sizeof(*cmdline);
-
         if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
                 efi_chunk_size = EFI_READ_CHUNK_SIZE;
 
         alloc_addr = alloc_size = 0;
+
+#ifdef CONFIG_CMDLINE
+       if (IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
+           IS_ENABLED(CONFIG_CMDLINE_OVERRIDE) ||
+           (again = (IS_ENABLED(CONFIG_X86) ||
+                     IS_ENABLED(CONFIG_CMDLINE_EXTEND)))) {
+               static const efi_char16_t builtin_cmdline[] = L""
 CONFIG_CMDLINE;
+
+               cmdline = builtin_cmdline;
+               cmdline_len = sizeof(builtin_cmdline);
+       } else
+#endif
+       {
+do_load_options:
+               cmdline = efi_table_attr(image, load_options);
+               cmdline_len = efi_table_attr(image, load_options_size);
+
+               efi_apply_loadoptions_quirk((const void **)&cmdline,
+                                           &cmdline_len);
+
+               again = false;
+       }
+       cmdline_len /= sizeof(*cmdline);
+
         do {
                 struct finfo fi;
                 unsigned long size;
drivers/firmware/efi/libstub/file.c +3 -0
@@ -290,6 +312,9 @@
                 efi_call_proto(volume, close);
         } while (offset > 0);
 
+       if (again)
+               goto do_load_options;
+
         *load_addr = alloc_addr;
         *load_size = alloc_size;
 
#10 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

In reply to Jonathan Marek
kernel test robot noticed the following build errors:

Hi Jonathan,

kernel test robot noticed the following build errors:

[auto build test ERROR on efi/next]
[also build test ERROR on linus/master v6.12-rc3 next-20241014] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Jonathan-Marek/efi-libstub-remove-uneccessary-cmdline_size-init-check/20241012-065337
base:   https://git.kernel.org/pub/scm/linux/kernel/git/efi/efi.git next
patch link:    https://lore.kernel.org/r/20241011224812.25763-3-jonathan%40marek.ca
patch subject: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options
config: x86_64-defconfig (https://download.01.org/0day-ci/archive/20241015/[email protected]/config)
compiler: gcc-11 (Debian 11.3.0-12) 11.3.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20241015/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

drivers/firmware/efi/libstub/file.c: In function 'handle_cmdline_files':

>> drivers/firmware/efi/libstub/file.c:218:38: error: expected ';' before 'CONFIG_CMDLINE'
     218 |                         cmdline = L"" CONFIG_CMDLINE;
         |                                      ^~~~~~~~~~~~~~~
         |                                      ;
   In file included from include/linux/string.h:6,
                    from include/linux/efi.h:16,
                    from drivers/firmware/efi/libstub/file.c:10:
>> drivers/firmware/efi/libstub/file.c:219:54: error: expected ')' before 'CONFIG_CMDLINE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^~~~~~~~~~~~~~
   include/linux/array_size.h:11:33: note: in definition of macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                 ^~~
   include/linux/array_size.h:11:32: note: to match this '('
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                ^
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~
>> drivers/firmware/efi/libstub/file.c:219:54: error: expected ')' before 'CONFIG_CMDLINE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^~~~~~~~~~~~~~
   include/linux/array_size.h:11:48: note: in definition of macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                ^~~
   include/linux/array_size.h:11:47: note: to match this '('
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                               ^
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~
   In file included from include/linux/init.h:5,
                    from include/linux/efi.h:15,
                    from drivers/firmware/efi/libstub/file.c:10:
>> drivers/firmware/efi/libstub/file.c:219:54: error: expected ')' before 'CONFIG_CMDLINE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^~~~~~~~~~~~~~
   include/linux/build_bug.h:16:62: note: in definition of macro 'BUILD_BUG_ON_ZERO'
      16 | #define BUILD_BUG_ON_ZERO(e) ((int)(sizeof(struct { int:(-!!(e)); })))
         |                                                              ^
   include/linux/compiler.h:243:51: note: in expansion of macro '__same_type'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                   ^~~~~~~~~~~
   include/linux/array_size.h:11:59: note: in expansion of macro '__must_be_array'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^~~~~~~~~~~~~~~
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~
   include/linux/compiler.h:243:63: note: to match this '('
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                               ^
   include/linux/build_bug.h:16:62: note: in definition of macro 'BUILD_BUG_ON_ZERO'
      16 | #define BUILD_BUG_ON_ZERO(e) ((int)(sizeof(struct { int:(-!!(e)); })))
         |                                                              ^
   include/linux/compiler.h:243:51: note: in expansion of macro '__same_type'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                   ^~~~~~~~~~~
   include/linux/array_size.h:11:59: note: in expansion of macro '__must_be_array'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^~~~~~~~~~~~~~~
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~
>> drivers/firmware/efi/libstub/file.c:219:54: error: expected ')' before 'CONFIG_CMDLINE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^~~~~~~~~~~~~~
   include/linux/build_bug.h:16:62: note: in definition of macro 'BUILD_BUG_ON_ZERO'
      16 | #define BUILD_BUG_ON_ZERO(e) ((int)(sizeof(struct { int:(-!!(e)); })))
         |                                                              ^
   include/linux/compiler.h:243:51: note: in expansion of macro '__same_type'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                   ^~~~~~~~~~~
   include/linux/array_size.h:11:59: note: in expansion of macro '__must_be_array'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^~~~~~~~~~~~~~~
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~
   include/linux/compiler.h:243:69: note: to match this '('
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                                     ^
   include/linux/build_bug.h:16:62: note: in definition of macro 'BUILD_BUG_ON_ZERO'
      16 | #define BUILD_BUG_ON_ZERO(e) ((int)(sizeof(struct { int:(-!!(e)); })))
         |                                                              ^
   include/linux/compiler.h:243:51: note: in expansion of macro '__same_type'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                   ^~~~~~~~~~~
   include/linux/array_size.h:11:59: note: in expansion of macro '__must_be_array'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^~~~~~~~~~~~~~~
   drivers/firmware/efi/libstub/file.c:219:39: note: in expansion of macro 'ARRAY_SIZE'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^~~~~~~~~~

vim +218 drivers/firmware/efi/libstub/file.c

177
178 /*
179 * Check the cmdline for a LILO-style file= arguments.
180 *
181 * We only support loading a file from the same filesystem as
182 * the kernel image.
183 */
184 efi_status_t handle_cmdline_files(efi_loaded_image_t *image,

   185					  const efi_char16_t *optstr,
   186					  int optstr_size,
   187					  unsigned long soft_limit,
   188					  unsigned long hard_limit,
   189					  unsigned long *load_addr,
   190					  unsigned long *load_size)
   191	{
   192		const efi_char16_t *cmdline = efi_table_attr(image, load_options);
   193		u32 cmdline_len = efi_table_attr(image, load_options_size);
   194		unsigned long efi_chunk_size = ULONG_MAX;
   195		efi_file_protocol_t *volume = NULL;
   196		efi_file_protocol_t *file;
   197		unsigned long alloc_addr;
   198		unsigned long alloc_size;
   199		efi_status_t status;
   200		int offset;
   201	
   202		if (!load_addr || !load_size)
   203			return EFI_INVALID_PARAMETER;
   204	
   205		efi_apply_loadoptions_quirk((const void **)&cmdline, &cmdline_len);
   206		cmdline_len /= sizeof(*cmdline);
   207	
   208		if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
   209			efi_chunk_size = EFI_READ_CHUNK_SIZE;
   210	
   211		if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
   212		    IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
   213		    cmdline_len == 0) {
   214			if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_len > 0) {
   215				/* both CONFIG_CMDLINE and load_options should be used */
   216				efi_warn("ignoring %ls from CONFIG_CMDLINE\n", optstr);
   217			} else {
 > 218				cmdline = L"" CONFIG_CMDLINE;
 > 219				cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#11 lore raw

Re: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options

In reply to Jonathan Marek
kernel test robot noticed the following build errors:

Hi Jonathan,

kernel test robot noticed the following build errors:

[auto build test ERROR on efi/next]
[also build test ERROR on linus/master v6.12-rc3 next-20241014] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Jonathan-Marek/efi-libstub-remove-uneccessary-cmdline_size-init-check/20241012-065337
base:   https://git.kernel.org/pub/scm/linux/kernel/git/efi/efi.git next
patch link:    https://lore.kernel.org/r/20241011224812.25763-3-jonathan%40marek.ca
patch subject: [PATCH 3/3] efi/libstub: consider CONFIG_CMDLINE for initrd= and dtb= options
config: i386-defconfig (https://download.01.org/0day-ci/archive/20241015/[email protected]/config)
compiler: clang version 18.1.8 (https://github.com/llvm/llvm-project 3b5b5c1ec4a3095ab096dd780e84d7ab81f3d7ff)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20241015/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

>> drivers/firmware/efi/libstub/file.c:218:17: error: expected ';' after expression
     218 |                         cmdline = L"" CONFIG_CMDLINE;
         |                                      ^
         |                                      ;
>> drivers/firmware/efi/libstub/file.c:218:18: error: use of undeclared identifier 'CONFIG_CMDLINE'
     218 |                         cmdline = L"" CONFIG_CMDLINE;
         |                                       ^
>> drivers/firmware/efi/libstub/file.c:219:33: error: expected ')'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^
   drivers/firmware/efi/libstub/file.c:219:18: note: to match this '('
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^
   include/linux/array_size.h:11:32: note: expanded from macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                ^
>> drivers/firmware/efi/libstub/file.c:219:33: error: expected ')'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^
   drivers/firmware/efi/libstub/file.c:219:18: note: to match this '('
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^
   include/linux/array_size.h:11:47: note: expanded from macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                               ^
>> drivers/firmware/efi/libstub/file.c:219:33: error: expected ')'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^
   drivers/firmware/efi/libstub/file.c:219:18: note: to match this '('
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^
   include/linux/array_size.h:11:59: note: expanded from macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^
   include/linux/compiler.h:243:58: note: expanded from macro '__must_be_array'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                               ^
>> drivers/firmware/efi/libstub/file.c:219:33: error: expected ')'
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                                      ^
   drivers/firmware/efi/libstub/file.c:219:18: note: to match this '('
     219 |                         cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
         |                                       ^
   include/linux/array_size.h:11:59: note: expanded from macro 'ARRAY_SIZE'
      11 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
         |                                                           ^
   include/linux/compiler.h:243:64: note: expanded from macro '__must_be_array'
     243 | #define __must_be_array(a)      BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0]))
         |                                                                     ^

6 errors generated.

vim +218 drivers/firmware/efi/libstub/file.c

177
178 /*
179 * Check the cmdline for a LILO-style file= arguments.
180 *
181 * We only support loading a file from the same filesystem as
182 * the kernel image.
183 */
184 efi_status_t handle_cmdline_files(efi_loaded_image_t *image,

   185					  const efi_char16_t *optstr,
   186					  int optstr_size,
   187					  unsigned long soft_limit,
   188					  unsigned long hard_limit,
   189					  unsigned long *load_addr,
   190					  unsigned long *load_size)
   191	{
   192		const efi_char16_t *cmdline = efi_table_attr(image, load_options);
   193		u32 cmdline_len = efi_table_attr(image, load_options_size);
   194		unsigned long efi_chunk_size = ULONG_MAX;
   195		efi_file_protocol_t *volume = NULL;
   196		efi_file_protocol_t *file;
   197		unsigned long alloc_addr;
   198		unsigned long alloc_size;
   199		efi_status_t status;
   200		int offset;
   201	
   202		if (!load_addr || !load_size)
   203			return EFI_INVALID_PARAMETER;
   204	
   205		efi_apply_loadoptions_quirk((const void **)&cmdline, &cmdline_len);
   206		cmdline_len /= sizeof(*cmdline);
   207	
   208		if (IS_ENABLED(CONFIG_X86) && !efi_nochunk)
   209			efi_chunk_size = EFI_READ_CHUNK_SIZE;
   210	
   211		if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
   212		    IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
   213		    cmdline_len == 0) {
   214			if (!IS_ENABLED(CONFIG_CMDLINE_FORCE) && cmdline_len > 0) {
   215				/* both CONFIG_CMDLINE and load_options should be used */
   216				efi_warn("ignoring %ls from CONFIG_CMDLINE\n", optstr);
   217			} else {
 > 218				cmdline = L"" CONFIG_CMDLINE;
 > 219				cmdline_len = ARRAY_SIZE(L"" CONFIG_CMDLINE) - 1;
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
maintainer
#12 lore raw
In reply to Jonathan Marek

Hi Jonathan,

Please use a cover letter when sending more than a single patch.

Quoted 18 lines · Jonathan Marek

efi_convert_cmdline() always returns a size of at least 1 because it counts the NUL terminator, so the "cmdline_size == 0" condition is not possible.

Change it to compare against 1 to get the intended behavior: to use CONFIG_CMDLINE when load_options_size is 0.

Fixes 60f38de7a8d4 ("efi/libstub: Unify command line param parsing")
---
 drivers/firmware/efi/libstub/efi-stub.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/firmware/efi/libstub/efi-stub.c +0 -0
index 958a680e0660d..709ae2d41a632 100644

drivers/firmware/efi/libstub/efi-stub.c efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr) · line 132

Quoted 5 lines
drivers/firmware/efi/libstub/efi-stub.c +1 -1
@@ -129,7 +129,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 
        if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
            IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
-           cmdline_size == 0) {
+           cmdline_size == 1) {

I'd prefer it if we could keep the weirdness local to efi_convert_cmdline(). Would the below fix things too?

drivers/firmware/efi/libstub/efi-stub-helper.c +1 -3
@@ -395,9 +395,7 @@
                }
        }
 
-       options_bytes++;        /* NUL termination */
-
-       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes,
+       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes + 1,
                             (void **)&cmdline_addr);
        if (status != EFI_SUCCESS)
                return NULL;

Note that the only other caller of efi_convert_cmdline() in x86-stub.c ignores this value entirely.

#13 lore raw
Quoted 46 lines · Ard Biesheuvel

Hi Jonathan,

Please use a cover letter when sending more than a single patch.

Quoted 18 lines · Jonathan Marek

efi_convert_cmdline() always returns a size of at least 1 because it counts the NUL terminator, so the "cmdline_size == 0" condition is not possible.

Change it to compare against 1 to get the intended behavior: to use CONFIG_CMDLINE when load_options_size is 0.

Fixes 60f38de7a8d4 ("efi/libstub: Unify command line param parsing")
---
  drivers/firmware/efi/libstub/efi-stub.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/firmware/efi/libstub/efi-stub.c +0 -0
index 958a680e0660d..709ae2d41a632 100644

drivers/firmware/efi/libstub/efi-stub.c efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr) · line 132

Quoted 5 lines
drivers/firmware/efi/libstub/efi-stub.c +1 -1
@@ -129,7 +129,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 
         if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
             IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
-           cmdline_size == 0) {
+           cmdline_size == 1) {

I'd prefer it if we could keep the weirdness local to efi_convert_cmdline(). Would the below fix things too?

drivers/firmware/efi/libstub/efi-stub-helper.c +1 -3
@@ -395,9 +395,7 @@
                 }
         }
 
-       options_bytes++;        /* NUL termination */
-
-       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes,
+       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes + 1,
                              (void **)&cmdline_addr);
         if (status != EFI_SUCCESS)
                 return NULL;
Quoted 3 lines

Note that the only other caller of efi_convert_cmdline() in x86-stub.c ignores this value entirely.

Just changing this would just make things more broken, the following snprintf would remove the last character of the command line because it uses options_bytes.

Since this patch has a Fixes: tag, I wanted to make the fix as simple as possible. If you think comparing the size to 1 is "weird", the fix could instead check if cmdline[0] is non-NUL (or just strlen(cmdline)==0 if you don't like that either).

And then my followup cleanup patch can just remove the cmd_line_len argument from efi_convert_cmdline().

maintainer
#14 lore raw
Quoted 48 lines · Jonathan Marek
Quoted 46 lines · Ard Biesheuvel

Hi Jonathan,

Please use a cover letter when sending more than a single patch.

Quoted 18 lines · Jonathan Marek

efi_convert_cmdline() always returns a size of at least 1 because it counts the NUL terminator, so the "cmdline_size == 0" condition is not possible.

Change it to compare against 1 to get the intended behavior: to use CONFIG_CMDLINE when load_options_size is 0.

Fixes 60f38de7a8d4 ("efi/libstub: Unify command line param parsing")
---
  drivers/firmware/efi/libstub/efi-stub.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/firmware/efi/libstub/efi-stub.c +0 -0
index 958a680e0660d..709ae2d41a632 100644

drivers/firmware/efi/libstub/efi-stub.c efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr) · line 132

Quoted 5 lines
drivers/firmware/efi/libstub/efi-stub.c +1 -1
@@ -129,7 +129,7 @@ efi_status_t efi_handle_cmdline(efi_loaded_image_t *image, char **cmdline_ptr)
 
         if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
             IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
-           cmdline_size == 0) {
+           cmdline_size == 1) {

I'd prefer it if we could keep the weirdness local to efi_convert_cmdline(). Would the below fix things too?

drivers/firmware/efi/libstub/efi-stub-helper.c +1 -3
@@ -395,9 +395,7 @@
                 }
         }
 
-       options_bytes++;        /* NUL termination */
-
-       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes,
+       status = efi_bs_call(allocate_pool, EFI_LOADER_DATA, options_bytes + 1,
                              (void **)&cmdline_addr);
         if (status != EFI_SUCCESS)
                 return NULL;
Quoted 8 lines
Quoted 3 lines

Note that the only other caller of efi_convert_cmdline() in x86-stub.c ignores this value entirely.

Just changing this would just make things more broken, the following snprintf would remove the last character of the command line because it uses options_bytes.

Ugh, you're right. So just use options_bytes - 1 in the assignment then.

Quoted 5 lines

Since this patch has a Fixes: tag, I wanted to make the fix as simple as possible. If you think comparing the size to 1 is "weird", the fix could instead check if cmdline[0] is non-NUL (or just strlen(cmdline)==0 if you don't like that either).

Checking the value of the first byte is reasonable I think.

Quoted 2 lines

And then my followup cleanup patch can just remove the cmd_line_len argument from efi_convert_cmdline().

Yes that would be fine - it is not really useful in any case.