Skip to contentSkip to conversation

linux: implement LoadFile2 initrd loading

PATCH In review Ard Biesheuvel 2 participants 9 messages Oct 18, 2022 – Oct 21, 2022

lore
Share & tools
lore
#0 lore raw

v5 cover linux: implement LoadFile2 initrd loading

This implements the LoadFile2 initrd loading protocol, which is essentially a callback interface into the bootloader to load the initrd data into a caller provided buffer. This means the bootloader no longer has to contain any policy regarding where to load the initrd (which differs between architectures and kernel versions) and no longer has to manipulate arch specific data structures such as DT or struct bootparams to inform the OS where the initrd resides in memory. This is especially relevant for the upcoming LoongArch support, which does not use either DT or struct bootparams, and would have to rely on the initrd= command line interface, which is deprecated and of limited utility [0].

Sample output from booting a recent Linux/arm64 kernel:

grub> insmod part_msdos
grub> linux (hd0,msdos1)/Image
grub> initrd (hd0,msdos1)/initrd.img
grub> boot
EFI stub: Booting Linux Kernel...
EFI stub: EFI_RNG_PROTOCOL unavailable
EFI stub: Generating empty DTB
EFI stub: Loaded initrd from LINUX_EFI_INITRD_MEDIA_GUID device path EFI stub: Exiting boot services...

  [    0.000000] Booting Linux on physical CPU 0x0000000000 [0x411fd070]

Changes since v4:
- incorporate review feedback from Daniel regarding whitespace and other minor style deviations

Changes since v3:
- drop a couple of patches that have been merged independently in the meantime
- update patch #1 to read the PE image header offset from the file instead of using the harcoded offset
- add acks from Heinrich and Ilias

Changes since v2:
- incorporate some ancient feedback from Daniel that I never saw until today. (this is why I am sending two versions of the same series on the same day - apologies for the spam)

[0] The initrd= command line loader can only access files that reside on the same volume as the loaded image, which means GRUB would have to present this volume abstraction in order to serve the initrd file.

Cc Daniel Kiper <[email protected]>
Cc Leif Lindholm <[email protected]>
Cc Nikita Ermakov <[email protected]>
Cc Atish Patra <[email protected]>
Cc Huacai Chen <[email protected]>
Cc Heinrich Schuchardt <[email protected]>
Cc dann frazier <[email protected]>
Cc Julian Andres Klode <[email protected]>
Cc Ilias Apalodimas <[email protected]>

Ard Biesheuvel (6):
efi: move MS-DOS stub out of generic PE header definition linux/arm: unify ARM/arm64 vs Xen PE/COFF header handling linux/arm: account for COFF headers appearing at unexpected offsets efi/efinet: Don't close connections at fini_hw() time
efi: implement LoadFile2 initrd loading protocol for Linux
linux: ignore FDT unless we need to modify it

 grub-core/commands/efi/lsefi.c     |   1 +
 grub-core/kern/efi/efi.c           |   8 +-
 grub-core/loader/arm64/linux.c     | 173 ++++++++++++++++++--
 grub-core/loader/arm64/xen_boot.c  |  23 +--
 grub-core/loader/efi/fdt.c         |   7 +-
 grub-core/net/drivers/efi/efinet.c |  10 +-
 grub-core/net/net.c                |   2 +-
 include/grub/arm/linux.h           |   5 +
 include/grub/arm64/linux.h         |   2 +
 include/grub/efi/api.h             |  40 +++++
 include/grub/efi/efi.h             |   4 +-
 include/grub/efi/pe32.h            |  16 +-
 include/grub/net.h                 |   3 +-
 13 files changed, 246 insertions(+), 48 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH v5 1/6 efi: move MS-DOS stub out of generic PE header definition

The PE/COFF spec permits the COFF signature and file header to appear anywhere in the file, and the actual offset is recorded in 4 byte little endian field at offset 0x3c of the image.

When GRUB is emitted as a PE/COFF binary, we reuse the 128 byte MS-DOS stub (even for non-x86 architectures), putting the COFF signature and file header at offset 0x80. However, other PE/COFF images may use different values, and non-x86 Linux kernels use an offset of 0x40 instead.

So let's get rid of the grub_pe32_header struct from pe32.h, given that it does not represent anything defined by the PE/COFF spec. Instead, introduce a minimal struct grub_msdos_image_header type based on the PE/COFF spec's description of the image header, and use the offset recorded at file position 0x3c to discover the actual location of the PE signature and the COFF image header.

The remaining fields are moved into a struct grub_pe_image_header, which we will use later to access COFF header fields of arbitrary images (and which may therefore appear at different offsets)

---
 grub-core/kern/efi/efi.c |  8 ++++++--
 include/grub/efi/pe32.h  | 16 ++++++++++++----
 2 files changed, 18 insertions(+), 6 deletions(-)
grub-core/kern/efi/efi.c +6 -2
index e8a976a22f15..cf49d6357e00 100644
@@ -302,7 +302,8 @@ grub_addr_t
 grub_efi_modules_addr (void)
 {
   grub_efi_loaded_image_t *image;
-  struct grub_pe32_header *header;
+  struct grub_msdos_image_header *header;
+  struct grub_pe_image_header *pe_image_header;
   struct grub_pe32_coff_header *coff_header;
   struct grub_pe32_section_table *sections;
   struct grub_pe32_section_table *section;
@@ -314,7 +315,10 @@ grub_efi_modules_addr (void)
     return 0;
 
   header = image->image_base;
-  coff_header = &(header->coff_header);
+  pe_image_header
+    = (struct grub_pe_image_header *) ((char *) header
+                                       + header->pe_image_header_offset);
+  coff_header = &(pe_image_header->coff_header);
   sections
     = (struct grub_pe32_section_table *) ((char *) coff_header
 					  + sizeof (*coff_header)
include/grub/efi/pe32.h +12 -4
index 0ed8781f0376..98c4ff177e19 100644
@@ -48,6 +48,17 @@
 
 #define GRUB_PE32_MAGIC			0x5a4d
 
+struct grub_msdos_image_header
+{
+  /* This is always 'MZ'. (GRUB_PE32_MAGIC)  */
+  grub_uint16_t msdos_magic;
+
+  grub_uint16_t reserved[29];
+
+  /* The file offset of the PE image header. */
+  grub_uint32_t pe_image_header_offset;
+};
+
 /* According to the spec, the minimal alignment is 512 bytes...
    But some examples (such as EFI drivers in the Intel
    Sample Implementation) use 32 bytes (0x20) instead, and it seems
@@ -254,11 +265,8 @@ struct grub_pe32_section_table
 
 #define GRUB_PE32_SIGNATURE_SIZE 4
 
-struct grub_pe32_header
+struct grub_pe_image_header
 {
-  /* This should be filled in with GRUB_PE32_MSDOS_STUB.  */
-  grub_uint8_t msdos_stub[GRUB_PE32_MSDOS_STUB_SIZE];
-
   /* This is always PE\0\0.  */
   char signature[GRUB_PE32_SIGNATURE_SIZE];
 
Signature · 1 line
2.35.1
#5 lore raw

PATCH v5 5/6 efi: implement LoadFile2 initrd loading protocol for Linux

In reply to Ard Biesheuvel

Recent Linux kernels will invoke the LoadFile2 protocol installed on a well-known vendor media path to load the initrd if it is exposed by the firmware. Using this method is preferred for two reasons:
- the Linux kernel is in charge of allocating the memory, and so it can implement any placement policy it wants (given that these tend to change between kernel versions),
- it is no longer necessary to modify the device tree provided by the firmware.

So let's install this protocol when handling the 'initrd' command if such a recent kernel was detected (based on the PE/COFF image version), and defer loading the initrd contents until the point where the kernel invokes the LoadFile2 protocol.

---
 grub-core/commands/efi/lsefi.c |   1 +
 grub-core/loader/arm64/linux.c | 124 +++++++++++++++++++-
 include/grub/efi/api.h         |  40 +++++++
 3 files changed, 164 insertions(+), 1 deletion(-)
grub-core/commands/efi/lsefi.c +1 -0
index f46ba3b49384..c304d25ccdd6 100644
@@ -55,6 +55,7 @@ struct known_protocol
     { GRUB_EFI_ABSOLUTE_POINTER_PROTOCOL_GUID, "absolute pointer" },
     { GRUB_EFI_DRIVER_BINDING_PROTOCOL_GUID, "EFI driver binding" },
     { GRUB_EFI_LOAD_FILE_PROTOCOL_GUID, "load file" },
+    { GRUB_EFI_LOAD_FILE2_PROTOCOL_GUID, "load file2" },
     { GRUB_EFI_SIMPLE_FILE_SYSTEM_PROTOCOL_GUID, "simple FS" },
     { GRUB_EFI_TAPE_IO_PROTOCOL_GUID, "tape I/O" },
     { GRUB_EFI_UNICODE_COLLATION_PROTOCOL_GUID, "unicode collation" },
grub-core/loader/arm64/linux.c +123 -1

172 lines not on this page · raw message

include/grub/efi/api.h +40 -0
index b0d003aa1164..b1a7259b92c8 100644
@@ -149,6 +149,11 @@
     { 0x8E, 0x3F, 0x00, 0xA0, 0xC9, 0x69, 0x72, 0x3B } \
   }
 
+#define GRUB_EFI_LOAD_FILE2_PROTOCOL_GUID \
+  { 0x4006c0c1, 0xfcb3, 0x403e, \
+    { 0x99, 0x6d, 0x4a, 0x6c, 0x87, 0x24, 0xe0, 0x6d } \
+  }
+
 #define GRUB_EFI_SIMPLE_FILE_SYSTEM_PROTOCOL_GUID \
   { 0x0964e5b22, 0x6459, 0x11d2, \
     { 0x8e, 0x39, 0x00, 0xa0, 0xc9, 0x69, 0x72, 0x3b } \
@@ -364,6 +369,11 @@
     { 0x86, 0x2e, 0xc0, 0x1c, 0xdc, 0x29, 0x1f, 0x44 } \
   }
 
+#define LINUX_EFI_INITRD_MEDIA_GUID  \
+  { 0x5568e427, 0x68fc, 0x4f3d, \
+    { 0xac, 0x74, 0xca, 0x55, 0x52, 0x31, 0xcc, 0x68 } \
+  }
+
 struct grub_efi_sal_system_table
 {
   grub_uint32_t signature;
@@ -564,6 +574,20 @@ typedef grub_uint16_t grub_efi_char16_t;
 
 typedef grub_efi_uintn_t grub_efi_status_t;
 
+/*
+ * On x86, the EFI calling convention may deviate from the local one, so
+ * callback functions exposed to the firmware must carry the follow attribute
+ * annotation. (This includes protocols implemented by GRUB that are installed
+ * into the EFI protocol database.)
+ */
+#if defined(__i386__)
+#define __grub_efi_api			__attribute__((regparm(0)))
+#elif defined(__x86_64__)
+#define __grub_efi_api			__attribute__((ms_abi))
+#else
+#define __grub_efi_api
+#endif
+
 #define GRUB_EFI_ERROR_CODE(value)	\
   ((((grub_efi_status_t) 1) << (sizeof (grub_efi_status_t) * 8 - 1)) | (value))
 
@@ -1755,6 +1779,22 @@ struct grub_efi_rng_protocol
 };
 typedef struct grub_efi_rng_protocol grub_efi_rng_protocol_t;
 
+struct grub_efi_load_file2
+{
+  grub_efi_status_t (__grub_efi_api *load_file)(struct grub_efi_load_file2 *this,
+                                                grub_efi_device_path_t *file_path,
+                                                grub_efi_boolean_t boot_policy,
+                                                grub_efi_uintn_t *buffer_size,
+                                                void *buffer);
+};
+typedef struct grub_efi_load_file2 grub_efi_load_file2_t;
+
+struct initrd_media_device_path {
+  grub_efi_vendor_media_device_path_t  vendor;
+  grub_efi_device_path_t               end;
+} GRUB_PACKED;
+typedef struct initrd_media_device_path initrd_media_device_path_t;
+
 #if (GRUB_TARGET_SIZEOF_VOID_P == 4) || defined (__ia64__) \
   || defined (__aarch64__) || defined (__MINGW64__) || defined (__CYGWIN__) \
   || defined(__riscv)
Signature · 1 line
2.35.1
#6 lore raw

PATCH v5 6/6 linux: ignore FDT unless we need to modify it

In reply to Ard Biesheuvel

Now that we implemented support for the LoadFile2 protocol for initrd loading, there is no longer a need to pass the initrd parameters via the device tree. This means that when the LoadFile2 protocol is being used, there is no reason to update the device tree in the first place, and so we can ignore it entirely.

The only remaining reason to deal with the devicetree is if we are using the 'devicetree' command to load one from disk, so tweak the logic in grub_fdt_install() to take that into account.

---
 grub-core/loader/arm64/linux.c | 22 ++++++++++----------
 grub-core/loader/efi/fdt.c     |  7 +++++--
 2 files changed, 16 insertions(+), 13 deletions(-)
grub-core/loader/arm64/linux.c +11 -11
index 9a4522d2f7bf..9d0bacc85419 100644
@@ -134,21 +134,21 @@ finalize_params_linux (void)
 
   void *fdt;
 
-  fdt = grub_fdt_load (GRUB_EFI_LINUX_FDT_EXTRA_SPACE);
+  /* Set initrd info */
+  if (initrd_start && initrd_end > initrd_start)
+    {
+      fdt = grub_fdt_load (GRUB_EFI_LINUX_FDT_EXTRA_SPACE);
 
-  if (!fdt)
-    goto failure;
+      if (!fdt)
+       goto failure;
 
-  node = grub_fdt_find_subnode (fdt, 0, "chosen");
-  if (node < 0)
-    node = grub_fdt_add_subnode (fdt, 0, "chosen");
+      node = grub_fdt_find_subnode (fdt, 0, "chosen");
+      if (node < 0)
+       node = grub_fdt_add_subnode (fdt, 0, "chosen");
 
-  if (node < 1)
-    goto failure;
+      if (node < 1)
+       goto failure;
 
-  /* Set initrd info */
-  if (initrd_start && initrd_end > initrd_start)
-    {
       grub_dprintf ("linux", "Initrd @ %p-%p\n",
 		    (void *) initrd_start, (void *) initrd_end);
 
grub-core/loader/efi/fdt.c +5 -2
index 80d70887474a..35a3be2e5cda 100644
@@ -89,13 +89,16 @@ grub_fdt_install (void)
   static grub_efi_guid_t fdt_guid = GRUB_EFI_DEVICE_TREE_GUID;
   grub_efi_status_t status;
 
+  if (fdt == NULL && loaded_fdt == NULL)
+    return GRUB_ERR_NONE;
+
   b = grub_efi_system_table->boot_services;
-  status = b->install_configuration_table (&fdt_guid, fdt);
+  status = b->install_configuration_table (&fdt_guid, fdt ? fdt : loaded_fdt);
   if (status != GRUB_EFI_SUCCESS)
     return grub_error (GRUB_ERR_IO, "failed to install FDT");
 
   grub_dprintf ("fdt", "Installed/updated FDT configuration table @ %p\n",
-		fdt);
+		fdt ? fdt : loaded_fdt);
   return GRUB_ERR_NONE;
 }
 
Signature · 1 line
2.35.1
#3 lore raw

PATCH v5 3/6 linux/arm: account for COFF headers appearing at unexpected offsets

In reply to Ard Biesheuvel

The way we load the Linux and PE/COFF image headers depends on a fixed placement of the COFF header at offset 0x40 into the file. This is a reasonable default, given that this is where Linux emits it today. However, in order to comply with the PE/COFF spec, which allows this header to appear anywhere in the file, let's ensure that we read the header from where it actually appears in the file if it is not located at offset 0x40.

---
 grub-core/loader/arm64/linux.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
grub-core/loader/arm64/linux.c +15 -0
index 3733a69d6e36..754db84a96cf 100644
@@ -63,6 +63,21 @@ grub_arch_efi_linux_load_image_header (grub_file_t file,
   grub_dprintf ("linux", "UEFI stub kernel:\n");
   grub_dprintf ("linux", "PE/COFF header @ %08x\n", lh->hdr_offset);
 
+  /*
+   * The PE/COFF spec permits the COFF header to appear anywhere in the file, so
+   * we need to double check whether it was where we expected it, and if not, we
+   * must load it from the correct offset into the pe_image_header field of
+   * struct linux_arch_kernel_header.
+   */
+  if ((grub_uint8_t *) lh + lh->hdr_offset != (grub_uint8_t *) &lh->pe_image_header)
+    {
+      if (grub_file_seek (file, lh->hdr_offset) == (grub_off_t) -1
+          || grub_file_read (file, &lh->pe_image_header,
+                             sizeof (struct grub_pe_image_header))
+             != sizeof (struct grub_pe_image_header))
+        return grub_error (GRUB_ERR_FILE_READ_ERROR, "failed to read COFF image header");
+    }
+
   return GRUB_ERR_NONE;
 }
 
Signature · 1 line
2.35.1
#2 lore raw

PATCH v5 2/6 linux/arm: unify ARM/arm64 vs Xen PE/COFF header handling

In reply to Ard Biesheuvel

Xen has its own version of the image header, to account for the additional PE/COFF header fields. Since we are adding references to those in the shared EFI loader code, update the common definitions and drop the Xen specific one which no longer has a purpose.

Since in both cases, the call to grub_arch_efi_linux_check_image() is preceded by a load of the image header, let's move the load into that function, and rename it to grub_arch_efi_linux_load_image_header()/

---
 grub-core/loader/arm64/linux.c    | 12 +++++-----
 grub-core/loader/arm64/xen_boot.c | 23 ++++----------------
 include/grub/arm/linux.h          |  5 +++++
 include/grub/arm64/linux.h        |  2 ++
 include/grub/efi/efi.h            |  4 +++-
 5 files changed, 21 insertions(+), 25 deletions(-)
grub-core/loader/arm64/linux.c +7 -5
index b5b559c236e0..3733a69d6e36 100644
@@ -49,8 +49,13 @@ static grub_addr_t initrd_start;
 static grub_addr_t initrd_end;
 
 grub_err_t
-grub_arch_efi_linux_check_image (struct linux_arch_kernel_header * lh)
+grub_arch_efi_linux_load_image_header (grub_file_t file,
+                                      struct linux_arch_kernel_header * lh)
 {
+  grub_file_seek (file, 0);
+  if (grub_file_read (file, lh, sizeof (*lh)) < (grub_ssize_t) sizeof (*lh))
+    return grub_error(GRUB_ERR_FILE_READ_ERROR, "failed to read Linux image header");
+
   if ((lh->code0 & 0xffff) != GRUB_PE32_MAGIC)
     return grub_error (GRUB_ERR_NOT_IMPLEMENTED_YET,
 		       N_("plain image kernel not supported - rebuild with CONFIG_(U)EFI_STUB enabled"));
@@ -301,10 +306,7 @@ grub_cmd_linux (grub_command_t cmd __attribute__ ((unused)),
 
   kernel_size = grub_file_size (file);
 
-  if (grub_file_read (file, &lh, sizeof (lh)) < (long) sizeof (lh))
-    return grub_errno;
-
-  if (grub_arch_efi_linux_check_image (&lh) != GRUB_ERR_NONE)
+  if (grub_arch_efi_linux_load_image_header (file, &lh) != GRUB_ERR_NONE)
     goto fail;
 
   grub_loader_unset();
grub-core/loader/arm64/xen_boot.c +4 -19
index 22cc25eccd9d..763d87dcde5b 100644
@@ -31,7 +31,6 @@
 #include <grub/efi/efi.h>
 #include <grub/efi/fdtload.h>
 #include <grub/efi/memory.h>
-#include <grub/efi/pe32.h>	/* required by struct xen_hypervisor_header */
 #include <grub/i18n.h>
 #include <grub/lib/cmdline.h>
 
@@ -65,18 +64,6 @@ enum module_type
 };
 typedef enum module_type module_type_t;
 
-struct xen_hypervisor_header
-{
-  struct linux_arm64_kernel_header efi_head;
-
-  /* This is always PE\0\0.  */
-  grub_uint8_t signature[GRUB_PE32_SIGNATURE_SIZE];
-  /* The COFF file header.  */
-  struct grub_pe32_coff_header coff_header;
-  /* The Optional header.  */
-  struct grub_pe64_optional_header optional_header;
-};
-
 struct xen_boot_binary
 {
   struct xen_boot_binary *next;
@@ -452,7 +439,7 @@ static grub_err_t
 grub_cmd_xen_hypervisor (grub_command_t cmd __attribute__ ((unused)),
 			 int argc, char *argv[])
 {
-  struct xen_hypervisor_header sh;
+  struct linux_arm64_kernel_header lh;
   grub_file_t file = NULL;
 
   grub_dl_ref (my_mod);
@@ -467,10 +454,7 @@ grub_cmd_xen_hypervisor (grub_command_t cmd __attribute__ ((unused)),
   if (!file)
     goto fail;
 
-  if (grub_file_read (file, &sh, sizeof (sh)) != (long) sizeof (sh))
-    goto fail;
-  if (grub_arch_efi_linux_check_image
-      ((struct linux_arch_kernel_header *) &sh) != GRUB_ERR_NONE)
+  if (grub_arch_efi_linux_load_image_header (file, &lh) != GRUB_ERR_NONE)
     goto fail;
   grub_file_seek (file, 0);
 
@@ -484,7 +468,8 @@ grub_cmd_xen_hypervisor (grub_command_t cmd __attribute__ ((unused)),
     return grub_errno;
 
   xen_hypervisor->is_hypervisor = 1;
-  xen_hypervisor->align = (grub_size_t) sh.optional_header.section_alignment;
+  xen_hypervisor->align
+    = (grub_size_t) lh.pe_image_header.optional_header.section_alignment;
 
   xen_boot_binary_load (xen_hypervisor, file, argc, argv);
   if (grub_errno == GRUB_ERR_NONE)
include/grub/arm/linux.h +5 -0
index bcd5a7eb186e..f38e695b1408 100644
@@ -22,6 +22,8 @@
 
 #include "system.h"
 
+#include <grub/efi/pe32.h>
+
 #define GRUB_LINUX_ARM_MAGIC_SIGNATURE 0x016f2818
 
 struct linux_arm_kernel_header {
@@ -32,6 +34,9 @@ struct linux_arm_kernel_header {
   grub_uint32_t end;   /* _edata */
   grub_uint32_t reserved2[3];
   grub_uint32_t hdr_offset;
+#if defined GRUB_MACHINE_EFI
+  struct grub_pe_image_header pe_image_header;
+#endif
 };
 
 #if defined(__arm__)
include/grub/arm64/linux.h +2 -0
index 7e22b4ab6990..3da71a51256f 100644
@@ -20,6 +20,7 @@
 #define GRUB_ARM64_LINUX_HEADER 1
 
 #include <grub/types.h>
+#include <grub/efi/pe32.h>
 
 #define GRUB_LINUX_ARM64_MAGIC_SIGNATURE 0x644d5241 /* 'ARM\x64' */
 
@@ -36,6 +37,7 @@ struct linux_arm64_kernel_header
   grub_uint64_t res4;		/* reserved */
   grub_uint32_t magic;		/* Magic number, little endian, "ARM\x64" */
   grub_uint32_t hdr_offset;	/* Offset of PE/COFF header */
+  struct grub_pe_image_header pe_image_header;
 };
 
 #if defined(__aarch64__)
include/grub/efi/efi.h +3 -1
index eb2dfdfce9f8..e61272de5330 100644
@@ -102,7 +102,9 @@ extern void (*EXPORT_VAR(grub_efi_net_config)) (grub_efi_handle_t hnd,
 void *EXPORT_FUNC(grub_efi_get_firmware_fdt)(void);
 grub_err_t EXPORT_FUNC(grub_efi_get_ram_base)(grub_addr_t *);
 #include <grub/cpu/linux.h>
-grub_err_t grub_arch_efi_linux_check_image(struct linux_arch_kernel_header *lh);
+#include <grub/file.h>
+grub_err_t grub_arch_efi_linux_load_image_header(grub_file_t file,
+                                                struct linux_arch_kernel_header *lh);
 grub_err_t grub_arch_efi_linux_boot_image(grub_addr_t addr, grub_size_t size,
                                            char *args);
 #endif
Signature · 1 line
2.35.1
#4 lore raw

PATCH v5 4/6 efi/efinet: Don't close connections at fini_hw() time

In reply to Ard Biesheuvel

When GRUB runs on top of EFI firmware, it only has access to block and network device abstractions exposed by the firmware, and it is up to the firmware to quiesce the underlying hardware when exiting boot services and handing over to the OS.

This is especially important for network devices, to prevent incoming packets from being DMA'd straight into memory after the OS has taken over but before it has managed to reconfigure the network hardware.

GRUB handles this by means of the grub_net_fini_hw() preboot hook, which is executed before calling into the booted image. This means that all network devices disappear or become inoperable before the EFI stub executes on EFI targeted builds. This is problematic as it prevents the EFI stub from calling back into GRUB provided protocols such as LoadFile2 for the initrd, which we will provide in a subsequent patch.

So add a flag that indicates to the network core that EFI network devices should not be closed when grub_net_fini_hw() is called.

---
 grub-core/net/drivers/efi/efinet.c | 10 +++++++++-
 grub-core/net/net.c                |  2 +-
 include/grub/net.h                 |  3 ++-
 3 files changed, 12 insertions(+), 3 deletions(-)
grub-core/net/drivers/efi/efinet.c +9 -1
index 73343d26d9e1..5adf5f40f492 100644
@@ -320,7 +320,15 @@ grub_efinet_findcards (void)
 
       card->name = grub_xasprintf ("efinet%d", i++);
       card->driver = &efidriver;
-      card->flags = 0;
+      /*
+       * EFI network devices are abstract SNP protocol instances, and the
+       * firmware is in charge of ensuring that they will be torn down when the
+       * OS loader hands off to the OS proper. Closing them as part of the
+       * preboot cleanup is therefore unnecessary, and undesirable, as it
+       * prevents us from using the network connection in a protocal callback
+       * such as LoadFile2 for initrd loading.
+       */
+      card->flags = GRUB_NET_CARD_NO_CLOSE_ON_FINI_HW;
       card->default_address.type = GRUB_NET_LINK_LEVEL_PROTOCOL_ETHERNET;
       grub_memcpy (card->default_address.mac,
 		   net->mode->current_address,
grub-core/net/net.c +1 -1
index 064e7114e012..7046dc57890a 100644
@@ -1787,7 +1787,7 @@ grub_net_fini_hw (int noreturn __attribute__ ((unused)))
 {
   struct grub_net_card *card;
   FOR_NET_CARDS (card)
-    if (card->opened)
+    if (card->opened && !(card->flags & GRUB_NET_CARD_NO_CLOSE_ON_FINI_HW))
       {
 	if (card->driver->close)
 	  card->driver->close (card);
include/grub/net.h +2 -1
index a64a04cc80b1..79cba357ae6a 100644
@@ -64,7 +64,8 @@ typedef enum grub_net_interface_flags
 typedef enum grub_net_card_flags
   {
     GRUB_NET_CARD_HWADDRESS_IMMUTABLE = 1,
-    GRUB_NET_CARD_NO_MANUAL_INTERFACES = 2
+    GRUB_NET_CARD_NO_MANUAL_INTERFACES = 2,
+    GRUB_NET_CARD_NO_CLOSE_ON_FINI_HW = 4
   } grub_net_card_flags_t;
 
 struct grub_net_card;
Signature · 1 line
2.35.1
Reviewed-by on the v5 series
#7 lore raw
In reply to Ard Biesheuvel
Quoted 54 lines · Ard Biesheuvel

This implements the LoadFile2 initrd loading protocol, which is essentially a callback interface into the bootloader to load the initrd data into a caller provided buffer. This means the bootloader no longer has to contain any policy regarding where to load the initrd (which differs between architectures and kernel versions) and no longer has to manipulate arch specific data structures such as DT or struct bootparams to inform the OS where the initrd resides in memory. This is especially relevant for the upcoming LoongArch support, which does not use either DT or struct bootparams, and would have to rely on the initrd= command line interface, which is deprecated and of limited utility [0].

Sample output from booting a recent Linux/arm64 kernel:

grub> insmod part_msdos
grub> linux (hd0,msdos1)/Image
grub> initrd (hd0,msdos1)/initrd.img
grub> boot
EFI stub: Booting Linux Kernel...
EFI stub: EFI_RNG_PROTOCOL unavailable
EFI stub: Generating empty DTB
EFI stub: Loaded initrd from LINUX_EFI_INITRD_MEDIA_GUID device path EFI stub: Exiting boot services...

  [    0.000000] Booting Linux on physical CPU 0x0000000000 [0x411fd070]

Changes since v4:
- incorporate review feedback from Daniel regarding whitespace and other minor style deviations

Changes since v3:
- drop a couple of patches that have been merged independently in the meantime
- update patch #1 to read the PE image header offset from the file instead of using the harcoded offset
- add acks from Heinrich and Ilias

Changes since v2:
- incorporate some ancient feedback from Daniel that I never saw until today. (this is why I am sending two versions of the same series on the same day - apologies for the spam)

[0] The initrd= command line loader can only access files that reside on the same volume as the loaded image, which means GRUB would have to present this volume abstraction in order to serve the initrd file.

Cc Daniel Kiper <[email protected]>
Cc Leif Lindholm <[email protected]>
Cc Nikita Ermakov <[email protected]>
Cc Atish Patra <[email protected]>
Cc Huacai Chen <[email protected]>
Cc Heinrich Schuchardt <[email protected]>
Cc dann frazier <[email protected]>
Cc Julian Andres Klode <[email protected]>
Cc Ilias Apalodimas <[email protected]>
Quoted 7 lines

Ard Biesheuvel (6):
efi: move MS-DOS stub out of generic PE header definition linux/arm: unify ARM/arm64 vs Xen PE/COFF header handling linux/arm: account for COFF headers appearing at unexpected offsets efi/efinet: Don't close connections at fini_hw() time
efi: implement LoadFile2 initrd loading protocol for Linux
linux: ignore FDT unless we need to modify it

Though one comment. Probably I would pull stdbool.h into include/grub/types.h in the patch #5. If I do not hear any objections I will do that for you before merging the patches.

Thank you for doing this work guys! And sorry it took so long to review all these patches...

Daniel

#8 lore raw
Quoted 64 lines · Daniel Kiper
Quoted 54 lines · Ard Biesheuvel

This implements the LoadFile2 initrd loading protocol, which is essentially a callback interface into the bootloader to load the initrd data into a caller provided buffer. This means the bootloader no longer has to contain any policy regarding where to load the initrd (which differs between architectures and kernel versions) and no longer has to manipulate arch specific data structures such as DT or struct bootparams to inform the OS where the initrd resides in memory. This is especially relevant for the upcoming LoongArch support, which does not use either DT or struct bootparams, and would have to rely on the initrd= command line interface, which is deprecated and of limited utility [0].

Sample output from booting a recent Linux/arm64 kernel:

grub> insmod part_msdos
grub> linux (hd0,msdos1)/Image
grub> initrd (hd0,msdos1)/initrd.img
grub> boot
EFI stub: Booting Linux Kernel...
EFI stub: EFI_RNG_PROTOCOL unavailable
EFI stub: Generating empty DTB
EFI stub: Loaded initrd from LINUX_EFI_INITRD_MEDIA_GUID device path EFI stub: Exiting boot services...

  [    0.000000] Booting Linux on physical CPU 0x0000000000 [0x411fd070]

Changes since v4:
- incorporate review feedback from Daniel regarding whitespace and other minor style deviations

Changes since v3:
- drop a couple of patches that have been merged independently in the meantime
- update patch #1 to read the PE image header offset from the file instead of using the harcoded offset
- add acks from Heinrich and Ilias

Changes since v2:
- incorporate some ancient feedback from Daniel that I never saw until today. (this is why I am sending two versions of the same series on the same day - apologies for the spam)

[0] The initrd= command line loader can only access files that reside on the same volume as the loaded image, which means GRUB would have to present this volume abstraction in order to serve the initrd file.

Cc Daniel Kiper <[email protected]>
Cc Leif Lindholm <[email protected]>
Cc Nikita Ermakov <[email protected]>
Cc Atish Patra <[email protected]>
Cc Huacai Chen <[email protected]>
Cc Heinrich Schuchardt <[email protected]>
Cc dann frazier <[email protected]>
Cc Julian Andres Klode <[email protected]>
Cc Ilias Apalodimas <[email protected]>
Quoted 7 lines

Ard Biesheuvel (6):
efi: move MS-DOS stub out of generic PE header definition linux/arm: unify ARM/arm64 vs Xen PE/COFF header handling linux/arm: account for COFF headers appearing at unexpected offsets efi/efinet: Don't close connections at fini_hw() time
efi: implement LoadFile2 initrd loading protocol for Linux
linux: ignore FDT unless we need to modify it

Quoted 2 lines

Thanks

Quoted 4 lines

Though one comment. Probably I would pull stdbool.h into include/grub/types.h in the patch #5. If I do not hear any objections I will do that for you before merging the patches.

Sure

Quoted 3 lines

Thank you for doing this work guys! And sorry it took so long to review all these patches...

No worries