The grub_net_network_level_interface_unregister(), previously
implemented in a header, did not remove the variables hooks that
were registered in grub_net_network_level_interface_register().
Fix this by implementing the same logic used to register the
variables and move the function into the grub-core/net/net.c.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The grub_disk_read() may trigger other disk reads, e.g. via loopbacks.
This may lead to very deep recursion which can corrupt the heap. So, fix
the issue by limiting reads depth.
Reported-by: B Horn <b@horn.uk>
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
It was possible to delete a loopback while there were still references
to it. This led to an exploitable use-after-free.
Fixed by implementing a reference counting in the grub_loopback struct.
Reported-by: B Horn <b@horn.uk>
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The GRUB may use TPM to verify the integrity of boot components and the
result can determine whether a previously sealed key can be released. If
everything checks out, showing nothing has been tampered with, the key
is released and GRUB unlocks the encrypted root partition for the next
stage of booting.
However, the liberal Command Line Interface (CLI) can be misused by
anyone in this case to access files in the encrypted partition one way
or another. Despite efforts to keep the CLI secure by preventing utility
command output from leaking file content, many techniques in the wild
could still be used to exploit the CLI, enabling attacks or learning
methods to attack. It's nearly impossible to account for all scenarios
where a hack could be applied.
Therefore, to mitigate potential misuse of the CLI after the root device
has been successfully unlocked via TPM, the user should be required to
authenticate using the LUKS password. This added layer of security
ensures that only authorized users can access the CLI reducing the risk
of exploitation or unauthorized access to the encrypted partition.
Fixes: CVE-2024-49504
Signed-off-by: Michael Chang <mchang@suse.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The grub_file_open() and grub_file_close() should be the only places
that allow a reference to a filesystem to stay open. So, add grub_dl_t
to grub_fs_t and set this in the GRUB_MOD_INIT() for each filesystem to
avoid issues when filesystems forget to do it themselves or do not track
their own references, e.g. squash4.
The fs_label(), fs_uuid(), fs_mtime() and fs_read() should all ref and
unref in the same function but it is essentially redundant in GRUB
single threaded model.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
It was possible to read OOB when an attribute had a size that exceeded
the allocated buffer. This resolves that by making sure all attributes
that get read are fully in the allocated space by implementing
a function to validate them.
Defining the offsets in include/grub/ntfs.h but they are only used in
the validation function and not across the rest of the NTFS code.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Right now to access the next attribute the code reads the length of the
current attribute and adds that to the current pointer. This is error
prone as bounds checking needs to be performed all over the place. So,
implement a helper and ensure its used across find_attr() and read_attr().
This commit does *not* implement full bounds checking. It is just the
preparation work for this to be added into the helper.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The end of the attribute buffer should be stored alongside the rest of
the attribute struct as right now it is not possible to implement bounds
checking when accessing attributes sequentially.
This is done via:
- updating init_attr() to set at->end and check is is not initially out of bounds,
- implementing checks as init_attr() had its type change in its callers,
- updating the value of at->end when needed.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
grub_strlcpy() acts the same way as strlcpy() does on most *NIX,
returning the length of src and ensuring dest is always NUL
terminated except when size is 0.
Signed-off-by: B Horn <b@horn.uk>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Move tpm_get_tpm_version() into grub_ieee1275_tpm_init() and invalidate
grub_ieee1275_tpm_ihandle in case no TPM 2 could be detected. Try the
initialization only once so that grub_tpm_present() will always return
the same result. Use the grub_ieee1275_tpm_ihandle as indicator for an
available TPM instead of grub_ieee1275_tpm_version, which can now be
removed.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Move common initialization functions from the ibmvtpm driver module into
tcg2.c that will be moved into the new TCG2 driver in a subsequent patch.
Make the functions available to the ibmvtpm driver as public functions
and variables.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Consolidate repeated definitions of IEEE1275_IHANDLE_INVALID that are cast
to the type grub_ieee1275_ihandle_t. On the occasion add "GRUB_" prefix to
the constant name.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
As a preparation to test tpm2_key_protector with grub-emu, the new
option, --tpm-device, is introduced to specify the TPM device for
grub-emu so that grub-emu can access an emulated TPM device from
the host.
Since grub-emu can directly access the device on host, it's easy to
implement the essential TCG2 command submission function with the
read/write functions and enable tpm2_key_protector module for grub-emu,
so that we can further test TPM2 key unsealing with grub-emu.
Signed-off-by: Gary Lin <glin@suse.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Tested-by: Stefan Berger <stefanb@linux.ibm.com>
Add a new parameter to cryptomount to support the key protectors framework: -P.
The parameter is used to automatically retrieve a key from specified key
protectors. The parameter may be repeated to specify any number of key
protectors. These are tried in order until one provides a usable key for any
given disk.
Signed-off-by: Hernan Gatta <hegatta@linux.microsoft.com>
Signed-off-by: Michael Chang <mchang@suse.com>
Signed-off-by: Gary Lin <glin@suse.com>
Reviewed-by: Glenn Washburn <development@efficientek.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Tested-by: Stefan Berger <stefanb@linux.ibm.com>
A key protector encapsulates functionality to retrieve an unlocking key
for a fully-encrypted disk from a specific source. A key protector
module registers itself with the key protectors framework when it is
loaded and unregisters when unloaded. Additionally, a key protector may
accept parameters that describe how it should operate.
The key protectors framework, besides offering registration and
unregistration functions, also offers a one-stop routine for finding and
invoking a key protector by name. If a key protector with the specified
name exists and if an unlocking key is successfully retrieved by it, the
function returns to the caller the retrieved key and its length.
Cc: Vladimir Serbinenko <phcoder@gmail.com>
Signed-off-by: Hernan Gatta <hegatta@linux.microsoft.com>
Signed-off-by: Gary Lin <glin@suse.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Tested-by: Stefan Berger <stefanb@linux.ibm.com>
In order to store the SBAT data we create a new ELF note. The string
".sbat", zero-padded to 4 byte alignment, shall be entered in the name
field. The string "SBAT"'s ASCII values, 0x53424154, should be entered
in the type field.
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Sudhakar Kuppusamy <sudhakar@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
For NX the GRUB binary has to announce that it is compatible with the
NX feature. This implies that when loading the executable GRUB image
several attributes are true:
- the binary doesn't need an executable stack,
- the binary doesn't need sections to be both executable and writable,
- the binary knows how to use the EFI Memory Attributes Protocol on code
it is loading.
This patch:
- adds a definition for the PE DLL Characteristics flag GRUB_PE32_NX_COMPAT,
- changes grub-mkimage to set that flag.
Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Jan Setje-Eilers <jan.setjeeilers@oracle.com>
Signed-off-by: Mate Kukri <mate.kukri@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
For NX we need to set the page access permission attributes for write
and execute permissions. This patch adds two new primitives, grub_set_mem_attrs()
and grub_clear_mem_attrs(), and associated constants definitions used
for that purpose. For most platforms it adds a dummy implementation.
On EFI platforms it implements the primitives using the EFI Memory
Attribute Protocol, defined in UEFI 2.10 specification.
Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Jan Setje-Eilers <jan.setjeeilers@oracle.com>
Signed-off-by: Mate Kukri <mate.kukri@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Currently .module_license is set writable, that is, the section has the
SHF_WRITE flag set, in the module's ELF headers. This probably never
actually matters but it can't possibly be correct. The patch sets that
data as "const" which causes that flag not to be set.
Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Jan Setje-Eilers <jan.setjeeilers@oracle.com>
Signed-off-by: Mate Kukri <mate.kukri@canonical.com>
Reviewed-By: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This is an x86-specific thing and should be available globally.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Signed-off-by: Sergii Dmytruk <sergii.dmytruk@3mdeb.com>
Reviewed-by: Ross Philipson <ross.philipson@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This fixes naming inconsistency that goes against coding style as well
as helps to avoid potential conflicts and confusion as this constant is
used in multiple places.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Signed-off-by: Sergii Dmytruk <sergii.dmytruk@3mdeb.com>
Reviewed-by: Ross Philipson <ross.philipson@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Currently rdmsr and wrmsr commands have own MSR support detection code.
This code is the same. So, it is duplicated. Additionally, this code
cannot be reused by others. Hence, extract this code to a function and
make it public. By the way, improve a code a bit.
Additionally, use GRUB_ERR_BAD_DEVICE instead of GRUB_ERR_BUG to signal
an error because errors encountered by this new routine are not bugs.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Signed-off-by: Sergii Dmytruk <sergii.dmytruk@3mdeb.com>
Reviewed-by: Ross Philipson <ross.philipson@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Use more obvious names which match corresponding instructions:
* grub_msr_read() => grub_rdmsr(),
* grub_msr_write() => grub_wrmsr().
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Signed-off-by: Sergii Dmytruk <sergii.dmytruk@3mdeb.com>
Reviewed-by: Ross Philipson <ross.philipson@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
It does not make sense to have separate headers for individual static
functions. So, make one common place to store them.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Signed-off-by: Sergii Dmytruk <sergii.dmytruk@3mdeb.com>
Reviewed-by: Ross Philipson <ross.philipson@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Module structure has natural alignment of 4. Respect it explicitly
rather than relying on the fact that _end is usually aligned.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Current code works only if package matches binary name transformation rules.
It's often true but is not guaranteed.
Fixes: https://savannah.gnu.org/bugs/?64410
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Type 0x01 was introduced with the ACPI DBGP table and type 0x12 was introduced
with the ACPI DBG2 table. Type 0x12 is used by the ACPI SPCR table on recent
AWS bare-metal instances (c6i/c7i). Also give each debug type a proper name.
Signed-off-by: Udo Steinberg <udo@hypervisor.org>
Reviewed-by: Glenn Washburn <development@efficientek.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Without them, e.g., 0x80LL on 64-bit target is 32-bit byte-swapped to
0xffffffff80000000 instead of correct 0x80000000.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Add functionality to disable command line interface access and editing of GRUB
menu entries if GRUB image is built with --disable-cli.
Signed-off-by: Alec Brown <alec.r.brown@oracle.com>
Reviewed-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The following EROFS patch will use this helper to handle
ALIGN_UP() overflow.
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The "ground truth" stack protector cookie value is kept in a global
variable, and loaded in every function prologue and epilogue to store
it into resp. compare it with the stack slot holding the cookie.
If the comparison fails, the program aborts, and this might occur
spuriously when the global variable changes values between the entry and
exit of a function. This implies that assigning the global variable at
boot should not involve any instrumented function calls, unless special
care is taken to ensure that the live call stack is synchronized, which
is non-trivial.
So avoid any function calls, including grub_memcpy(), which is
unnecessary given that the stack cookie is always a suitably aligned
variable of the native word size.
While at it, leave the last byte 0x0 to avoid inadvertent unbounded
strings on the stack.
Note that the use of __attribute__((optimize)) is described as
unsuitable for production use in the GCC documentation, so let's drop
this as well now that it is no longer needed.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Glenn Washburn <development@efficientek.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Otherwise the GRUB cannot start due to missing symbols when stack
protector is enabled on EFI platforms.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Vladimir Serbinenko <phcoder@gmail.com>
The CMOS actually exists on most EFI platforms and in some cases is used to
store useful data that makes it justifiable for GRUB to read/write it.
As for date and time keep using EFI API and not CMOS one.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
No alignment is guaranteed and in fact on my IA-64 SAPIC is aligned
to 4 bytes instead of 8 and causes a trap. It affects only rarely used
lsacpi command and so went unnoticed.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
NetBSD uses slightly different function names for the same functions.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Introduce flags to identify PowerVM and KVM on Power and set them where
each type of host has been detected.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
Rename regions_claim() to grub_regions_claim() to make it available for
memory allocation. The ieee1275 loader will use this function on PowerVM
and KVM on Power and thus avoid usage of memory that it is not allowed
to use.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
Add support for memory alignment requirements and adjust a candidate
address to it before checking whether the block is large enough. This
must be done in this order since the alignment adjustment can make
a block smaller than what was requested.
None of the current callers has memory alignment requirements but the
ieee1275 loader for kernel and initrd will use it to convey them.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
Return the allocated address of the memory block in the request structure
if a memory allocation was actually done. Leave the address untouched
otherwise. This enables a caller who wants to use the allocated memory
directly, rather than adding the memory to the heap, to see where memory
was allocated. None of the current callers need this but the converted
ieee1275 loader will make use of it.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
Let the regions_claim() request structure's init_region determine whether
to call grub_mm_init_region() on it. This allows for adding memory to
GRUB's memory heap if init_region is set to true, or direct usage of the
memory otherwise. Set all current callers' init_region to true since they
want to add memory regions to GRUB's heap.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
The regions_claim() function limits the allocation of memory regions
by excluding certain memory areas from being used by GRUB. This for
example includes a gap between 640MB and 768MB as well as an upper
limit beyond which no memory may be used when an fadump is present.
However, the ieee1275 loader for kernel and initrd currently does not
use regions_claim() for memory allocation on PowerVM and KVM on Power
and therefore may allocate memory in those areas that it should not use.
To make the regions_claim() function more flexible and ultimately usable
for the ieee1275 loader, introduce a request structure to pass various
parameters to the regions_claim() function that describe the properties
of requested memory chunks. In a first step, move the total and flags
variables into this structure.
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Pavithra Prakash <pavrampu@in.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carolyn Scherrer <cpscherr@us.ibm.com>
Cc: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Cc: Sourabh Jain <sourabhjain@linux.ibm.com>
On ia64 alignment requirements are strict. When we pass a pointer to
UUID it needs to be at least 4-byte aligned or EFI will crash.
On the other hand in device path there is no padding for UUID, so we
need 2 types in one formor another. Make 4-byte aligned and unaligned types
The code is structured in a way to accept unaligned inputs
in most cases and supply 4-byte aligned outputs.
Efiemu case is a bit ugly because there inputs and outputs are
reversed and so we need careful casts to account for this
inversion.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
gpt_partition contains grub_guid. We need to decide whether the whole
structure is unaligned and then we need to use packed_guid. But we never
have unaligned part entries as we read them in an aligned buffer from disk.
Hence just make it all aligned.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
We do table search in many places doing exactly the same algorithm.
The only minor variance in users is which table is used if several entries
are present. As specification mandates uniqueness and even if it ever isn't,
first entry is good enough, unify this code and always use the first entry.
Signed-off-by: Vladimir Serbinenko <phcoder@gmail.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Current code imposes limitations on the amount of sectors read in
a single call according to CHS layout of the disk even in LBA
read mode. There's no need to obey CHS layout restrictions for
LBA reads on LBA disks. It only slows down booting process.
See: https://lore.kernel.org/grub-devel/d42a11fa-2a59-b5e7-08b1-d2c60444bb99@valdikss.org.ru/
Signed-off-by: ValdikSS <iam@valdikss.org.ru>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Recent mixed-mode Linux kernels, i.e., v4.0 or newer, can access EFI
runtime services at OS runtime even when the OS was not entered via the
EFI stub. This is because, instead of reverting back to the firmware's
segment selectors, GDTs and IDTs, the 64-bit kernel simply calls 32-bit
runtime services using compatibility mode, i.e., the same mode used for
32-bit user space, without taking down all interrupt handling, exception
handling, etc.
This means that GRUB's legacy x86 boot mode is sufficient to make use of
this: 32-bit i686 builds of GRUB can already boot 64-bit kernels in EFI
enlightened mode, but without going via the EFI stub, and provide all
the metadata that the OS needs to map the EFI runtime regions and call
EFI runtime services successfully.
It does mean that GRUB should not attempt to invoke the firmware's
LoadImage()/StartImage() methods on kernel builds that it knows cannot
be started natively. So, add a check for this in the native EFI boot
path and fall back to legacy x86 mode in such cases.
Note that in the general case, booting non-native images of the same
native word size, e.g., x64 EFI apps on arm64 firmware, might be
supported by means of emulation. So, let's only disallow images that use
a non-native word size. This will also permit booting i686 kernels on
x86_64 builds, although without access to runtime services, as this is
not supported by Linux.
This change on top of 2.12-rc1 is sufficient to boot ordinary Linux
mixed mode builds and get full access to the EFI runtime services.
Cc: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Steve McIntyre <steve@einval.com>
Cc: Julian Andres Klode <julian.klode@canonical.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Dimitri John Ledkov <dimitri.ledkov@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The x86_64 Linux kernel can be booted in 32-bit mode, in which case the
startup code creates a set of preliminary page tables that map the first
4 GiB of physical memory 1:1 and enables paging. This is a prerequisite
for 64-bit execution and can therefore only be implemented in 32-bit code.
The x86_64 Linux kernel can also be booted in 64-bit mode directly: this
implies that paging is already enabled and it is the responsibility of
the bootloader to ensure that the active page tables cover the entire
loaded image, including its BSS space, the size of which is described in
the image's setup header.
Given that the EFI spec mandates execution in long mode for x86_64 and
stipulates that all system memory is mapped 1:1, the Linux/x86
requirements for 64-bit entry can be met trivially when booting on
x86_64 via EFI. So, enter via the 64-bit entry point in this case.
This involves inspecting the xloadflags field in the setup header to
check whether the 64-bit entry point is supported. This field was
introduced in Linux version v3.8 (early 2013).
This change ensures that all EFI firmware tables and other assets passed
by the firmware or bootloader in memory remain mapped and accessible
throughout the early startup code.
Avoiding the drop out of long mode will also be needed to support
upcoming CPU designs that no longer implement 32-bit mode at all
(as recently announced by Intel [0]).
[0] https://www.intel.com/content/www/us/en/developer/articles/technical/envisioning-future-simplified-architecture.html
Cc: Daniel Kiper <daniel.kiper@oracle.com>
Cc: Julian Andres Klode <julian.klode@canonical.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
