Kbuild builds init/built-in.a twice; first during the ordinary
directory descending, second from scripts/link-vmlinux.sh.
We do this because UTS_VERSION contains the build version and the
timestamp. We cannot update it during the normal directory traversal
since we do not yet know if we need to update vmlinux. UTS_VERSION is
temporarily calculated, but omitted from the update check. Otherwise,
vmlinux would be rebuilt every time.
When Kbuild results in running link-vmlinux.sh, it increments the
version number in the .version file and takes the timestamp at that
time to really fix UTS_VERSION.
However, updating the same file twice is a footgun. To avoid nasty
timestamp issues, all build artifacts that depend on init/built-in.a
are atomically generated in link-vmlinux.sh, where some of them do not
need rebuilding.
To fix this issue, this commit changes as follows:
[1] Split UTS_VERSION out to include/generated/utsversion.h from
include/generated/compile.h
include/generated/utsversion.h is generated just before the
vmlinux link. It is generated under include/generated/ because
some decompressors (s390, x86) use UTS_VERSION.
[2] Split init_uts_ns and linux_banner out to init/version-timestamp.c
from init/version.c
init_uts_ns and linux_banner contain UTS_VERSION. During the ordinary
directory descending, they are compiled with __weak and used to
determine if vmlinux needs relinking. Just before the vmlinux link,
they are compiled without __weak to embed the real version and
timestamp.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The early malloc() and free() implementation in include/linux/decompress/mm.h
(which is also included by the static decompressors) is static. This is
fine when the only thing interested in using malloc() is the decompression
code, but the x86 early boot environment may use malloc() in a couple places,
leading to a potential collision when the static copies of the available
memory region ("malloc_ptr") gets reset to the global "free_mem_ptr" value.
As it happened, the existing usage pattern was accidentally safe because each
user did 1 malloc() and 1 free() before returning and were not nested:
extract_kernel() (misc.c)
choose_random_location() (kaslr.c)
mem_avoid_init()
handle_mem_options()
malloc()
...
free()
...
parse_elf() (misc.c)
malloc()
...
free()
Once the future FGKASLR series is added, however, it will insert
additional malloc() calls local to fgkaslr.c in the middle of
parse_elf()'s malloc()/free() pair:
parse_elf() (misc.c)
malloc()
if (...) {
layout_randomized_image(output, &ehdr, phdrs);
malloc() <- boom
...
else
layout_image(output, &ehdr, phdrs);
free()
To avoid collisions, there must be a single implementation of malloc().
Adjust include/linux/decompress/mm.h so that visibility can be
controlled, provide prototypes in misc.h, and implement the functions in
misc.c. This also results in a small size savings:
$ size vmlinux.before vmlinux.after
text data bss dec hex filename
8842314 468 178320 9021102 89a6ae vmlinux.before
8842240 468 178320 9021028 89a664 vmlinux.after
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20211013175742.1197608-4-keescook@chromium.org
Commit
ca0e22d4f0 ("x86/boot/compressed/64: Always switch to own page table")
started using a new set of pagetables even without KASLR.
After that commit, initialize_identity_maps() is called before the
5-level paging variables are setup in choose_random_location(), which
will not work if 5-level paging is actually enabled.
Fix this by moving the initialization of __pgtable_l5_enabled,
pgdir_shift and ptrs_per_p4d into cleanup_trampoline(), which is called
immediately after the finalization of whether the kernel is executing
with 4- or 5-level paging. This will be earlier than anything that might
require those variables, and keeps the 4- vs 5-level paging code all in
one place.
Fixes: ca0e22d4f0 ("x86/boot/compressed/64: Always switch to own page table")
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Joerg Roedel <jroedel@suse.de>
Tested-by: Joerg Roedel <jroedel@suse.de>
Tested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Link: https://lkml.kernel.org/r/20201010191110.4060905-1-nivedita@alum.mit.edu
When booted through startup_64(), the kernel keeps running on the EFI
page table until the KASLR code sets up its own page table. Without
KASLR, the pre-decompression boot code never switches off the EFI page
table. Change that by unconditionally switching to a kernel-controlled
page table after relocation.
This makes sure the kernel can make changes to the mapping when
necessary, for example map pages unencrypted in SEV and SEV-ES guests.
Also, remove the debug_putstr() calls in initialize_identity_maps()
because the function now runs before console_init() is called.
[ bp: Massage commit message. ]
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lkml.kernel.org/r/20200907131613.12703-17-joro@8bytes.org
- Add support for zstd compressed kernel
- Define __DISABLE_EXPORTS in Makefile
- Remove __DISABLE_EXPORTS definition from kaslr.c
- Bump the heap size for zstd.
- Update the documentation.
Integrates the ZSTD decompression code to the x86 pre-boot code.
Zstandard requires slightly more memory during the kernel decompression
on x86 (192 KB vs 64 KB), and the memory usage is independent of the
window size.
__DISABLE_EXPORTS is now defined in the Makefile, which covers both
the existing use in kaslr.c, and the use needed by the zstd decompressor
in misc.c.
This patch has been boot tested with both a zstd and gzip compressed
kernel on i386 and x86_64 using buildroot and QEMU.
Additionally, this has been tested in production on x86_64 devices.
We saw a 2 second boot time reduction by switching kernel compression
from xz to zstd.
Signed-off-by: Nick Terrell <terrelln@fb.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20200730190841.2071656-7-nickrterrell@gmail.com
Drop unnecessary alignment of image_size to CONFIG_PHYSICAL_ALIGN in
find_random_virt_addr, it cannot change the result: the largest valid
slot is the largest n that satisfies
minimum + n * CONFIG_PHYSICAL_ALIGN + image_size <= KERNEL_IMAGE_SIZE
(since minimum is already aligned) and so n is equal to
(KERNEL_IMAGE_SIZE - minimum - image_size) / CONFIG_PHYSICAL_ALIGN
even if image_size is not aligned to CONFIG_PHYSICAL_ALIGN.
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20200728225722.67457-17-nivedita@alum.mit.edu
The number of slots can be 'unsigned int', since on 64-bit, the maximum
amount of memory is 2^52, the minimum alignment is 2^21, so the slot
number cannot be greater than 2^31. But in case future processors have
more than 52 physical address bits, make it 'unsigned long'.
The slot areas are limited by MAX_SLOT_AREA, currently 100. It is
indexed by an int, but the number of areas is stored as 'unsigned long'.
Change both to 'unsigned int' for consistency.
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20200728225722.67457-15-nivedita@alum.mit.edu
This check doesn't save anything. In the case when none of the
parameters are present, each strstr will scan args twice (once to find
the length and then for searching), six scans in total. Just going ahead
and parsing the arguments only requires three scans: strlen, memcpy, and
parsing. This will be the first malloc, so free will actually free up
the memory, so the check doesn't save heap space either.
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20200728225722.67457-14-nivedita@alum.mit.edu
Clip the start and end of the region to minimum and mem_limit prior to
the loop. region.start can only increase during the loop, so raising it
to minimum before the loop is enough.
A region that becomes empty due to this will get checked in
the first iteration of the loop.
Drop the check for overlap extending beyond the end of the region. This
will get checked in the next loop iteration anyway.
Rename end to region_end for symmetry with region.start.
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20200728225722.67457-10-nivedita@alum.mit.edu
On 64-bit, the kernel must be placed below MAXMEM (64TiB with 4-level
paging or 4PiB with 5-level paging). This is currently not enforced by
KASLR, which thus implicitly relies on physical memory being limited to
less than 64TiB.
On 32-bit, the limit is KERNEL_IMAGE_SIZE (512MiB). This is enforced by
special checks in __process_mem_region().
Initialize mem_limit to the maximum (depending on architecture), instead
of ULLONG_MAX, and make sure the command-line arguments can only
decrease it. This makes the enforcement explicit on 64-bit, and
eliminates the 32-bit specific checks to keep the kernel below 512M.
Check upfront to make sure the minimum address is below the limit before
doing any work.
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20200727230801.3468620-5-nivedita@alum.mit.edu
Pull ACPI updates from Rafael Wysocki:
"These update the ACPICA code in the kernel to upstream revision
20191018, add support for EFI specific purpose memory, update the ACPI
EC driver to make it work on systems with hardware-reduced ACPI,
improve ACPI-based device enumeration for some platforms, rework the
lid blacklist handling in the button driver and add more lid quirks to
it, unify ACPI _HID/_UID matching, fix assorted issues and clean up
the code and documentation.
Specifics:
- Update the ACPICA code in the kernel to upstream revision 20191018
including:
* Fixes for Clang warnings (Bob Moore)
* Fix for possible overflow in get_tick_count() (Bob Moore)
* Introduction of acpi_unload_table() (Bob Moore)
* Debugger and utilities updates (Erik Schmauss)
* Fix for unloading tables loaded via configfs (Nikolaus Voss)
- Add support for EFI specific purpose memory to optionally allow
either application-exclusive or core-kernel-mm managed access to
differentiated memory (Dan Williams)
- Fix and clean up processing of the HMAT table (Brice Goglin, Qian
Cai, Tao Xu)
- Update the ACPI EC driver to make it work on systems with
hardware-reduced ACPI (Daniel Drake)
- Always build in support for the Generic Event Device (GED) to allow
one kernel binary to work both on systems with full hardware ACPI
and hardware-reduced ACPI (Arjan van de Ven)
- Fix the table unload mechanism to unregister platform devices
created when the given table was loaded (Andy Shevchenko)
- Rework the lid blacklist handling in the button driver and add more
lid quirks to it (Hans de Goede)
- Improve ACPI-based device enumeration for some platforms based on
Intel BayTrail SoCs (Hans de Goede)
- Add an OpRegion driver for the Cherry Trail Crystal Cove PMIC and
prevent handlers from being registered for unhandled PMIC OpRegions
(Hans de Goede)
- Unify ACPI _HID/_UID matching (Andy Shevchenko)
- Clean up documentation and comments (Cao jin, James Pack, Kacper
Piwiński)"
* tag 'acpi-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (52 commits)
ACPI: OSI: Shoot duplicate word
ACPI: HMAT: use %u instead of %d to print u32 values
ACPI: NUMA: HMAT: fix a section mismatch
ACPI: HMAT: don't mix pxm and nid when setting memory target processor_pxm
ACPI: NUMA: HMAT: Register "soft reserved" memory as an "hmem" device
ACPI: NUMA: HMAT: Register HMAT at device_initcall level
device-dax: Add a driver for "hmem" devices
dax: Fix alloc_dax_region() compile warning
lib: Uplevel the pmem "region" ida to a global allocator
x86/efi: Add efi_fake_mem support for EFI_MEMORY_SP
arm/efi: EFI soft reservation to memblock
x86/efi: EFI soft reservation to E820 enumeration
efi: Common enable/disable infrastructure for EFI soft reservation
x86/efi: Push EFI_MEMMAP check into leaf routines
efi: Enumerate EFI_MEMORY_SP
ACPI: NUMA: Establish a new drivers/acpi/numa/ directory
ACPICA: Update version to 20191018
ACPICA: debugger: remove leading whitespaces when converting a string to a buffer
ACPICA: acpiexec: initialize all simple types and field units from user input
ACPICA: debugger: add field unit support for acpi_db_get_next_token
...
Given that EFI_MEMORY_SP is platform BIOS policy decision for marking
memory ranges as "reserved for a specific purpose" there will inevitably
be scenarios where the BIOS omits the attribute in situations where it
is desired. Unlike other attributes if the OS wants to reserve this
memory from the kernel the reservation needs to happen early in init. So
early, in fact, that it needs to happen before e820__memblock_setup()
which is a pre-requisite for efi_fake_memmap() that wants to allocate
memory for the updated table.
Introduce an x86 specific efi_fake_memmap_early() that can search for
attempts to set EFI_MEMORY_SP via efi_fake_mem and update the e820 table
accordingly.
The KASLR code that scans the command line looking for user-directed
memory reservations also needs to be updated to consider
"efi_fake_mem=nn@ss:0x40000" requests.
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
UEFI 2.8 defines an EFI_MEMORY_SP attribute bit to augment the
interpretation of the EFI Memory Types as "reserved for a specific
purpose".
The proposed Linux behavior for specific purpose memory is that it is
reserved for direct-access (device-dax) by default and not available for
any kernel usage, not even as an OOM fallback. Later, through udev
scripts or another init mechanism, these device-dax claimed ranges can
be reconfigured and hot-added to the available System-RAM with a unique
node identifier. This device-dax management scheme implements "soft" in
the "soft reserved" designation by allowing some or all of the
reservation to be recovered as typical memory. This policy can be
disabled at compile-time with CONFIG_EFI_SOFT_RESERVE=n, or runtime with
efi=nosoftreserve.
This patch introduces 2 new concepts at once given the entanglement
between early boot enumeration relative to memory that can optionally be
reserved from the kernel page allocator by default. The new concepts
are:
- E820_TYPE_SOFT_RESERVED: Upon detecting the EFI_MEMORY_SP
attribute on EFI_CONVENTIONAL memory, update the E820 map with this
new type. Only perform this classification if the
CONFIG_EFI_SOFT_RESERVE=y policy is enabled, otherwise treat it as
typical ram.
- IORES_DESC_SOFT_RESERVED: Add a new I/O resource descriptor for
a device driver to search iomem resources for application specific
memory. Teach the iomem code to identify such ranges as "Soft Reserved".
Note that the comment for do_add_efi_memmap() needed refreshing since it
seemed to imply that the efi map might overflow the e820 table, but that
is not an issue as of commit 7b6e4ba3cb "x86/boot/e820: Clean up the
E820_X_MAX definition" that removed the 128 entry limit for
e820__range_add().
A follow-on change integrates parsing of the ACPI HMAT to identify the
node and sub-range boundaries of EFI_MEMORY_SP designated memory. For
now, just identify and reserve memory of this type.
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reported-by: kbuild test robot <lkp@intel.com>
Reviewed-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When building randconfigs, one of the failures is:
ld: arch/x86/boot/compressed/kaslr.o: in function `choose_random_location':
kaslr.c:(.text+0xbf7): undefined reference to `count_immovable_mem_regions'
ld: kaslr.c:(.text+0xcbe): undefined reference to `immovable_mem'
make[2]: *** [arch/x86/boot/compressed/vmlinux] Error 1
because CONFIG_ACPI is not enabled in this particular .config but
CONFIG_MEMORY_HOTREMOVE is and count_immovable_mem_regions() is
unresolvable because it is defined in compressed/acpi.c which is the
compilation unit that depends on CONFIG_ACPI.
Add CONFIG_ACPI to the explicit dependencies for MEMORY_HOTREMOVE.
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Chao Fan <fanc.fnst@cn.fujitsu.com>
Cc: x86@kernel.org
Link: https://lkml.kernel.org/r/20190205131033.9564-1-bp@alien8.de
When KASLR is enabled then 1GB huge pages allocations might regress
sporadically.
To reproduce on a KVM guest with 4GB RAM:
- add the following options to the kernel command-line:
'default_hugepagesz=1G hugepagesz=1G hugepages=1'
- boot the guest and check number of 1GB pages reserved:
# grep HugePages_Total /proc/meminfo
- sporadically, every couple of bootups the output of this
command shows that when booting with "nokaslr" HugePages_Total is always 1,
while booting without "nokaslr" sometimes HugePages_Total is set as 0
(that is, reserving the 1GB page failed).
Note that you may need to boot a few times to trigger the issue,
because it's somewhat non-deterministic.
The root cause is that kernel may be put into the only good 1GB huge page
in the [0x40000000, 0x7fffffff] physical range randomly.
Below is the dmesg output snippet from the KVM guest. We can see that only
[0x40000000, 0x7fffffff] region is good 1GB huge page,
[0x100000000, 0x13fffffff] will be touched by the memblock top-down allocation:
[...] e820: BIOS-provided physical RAM map:
[...] BIOS-e820: [mem 0x0000000000000000-0x000000000009fbff] usable
[...] BIOS-e820: [mem 0x000000000009fc00-0x000000000009ffff] reserved
[...] BIOS-e820: [mem 0x00000000000f0000-0x00000000000fffff] reserved
[...] BIOS-e820: [mem 0x0000000000100000-0x00000000bffdffff] usable
[...] BIOS-e820: [mem 0x00000000bffe0000-0x00000000bfffffff] reserved
[...] BIOS-e820: [mem 0x00000000feffc000-0x00000000feffffff] reserved
[...] BIOS-e820: [mem 0x00000000fffc0000-0x00000000ffffffff] reserved
[...] BIOS-e820: [mem 0x0000000100000000-0x000000013fffffff] usable
Besides, on bare-metal machines with larger memory, one less 1GB huge page
might be available with KASLR enabled. That too is because the kernel
image might be randomized into those "good" 1GB huge pages.
To fix this, firstly parse the kernel command-line to get how many 1GB huge
pages are specified. Then try to skip the specified number of 1GB huge
pages when decide which memory region kernel can be randomized into.
Also change the name of handle_mem_memmap() as handle_mem_options()
since it handles not only 'mem=' and 'memmap=', but also 'hugepagesxxx' now.
Signed-off-by: Baoquan He <bhe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: douly.fnst@cn.fujitsu.com
Cc: fanc.fnst@cn.fujitsu.com
Cc: indou.takao@jp.fujitsu.com
Cc: keescook@chromium.org
Cc: lcapitulino@redhat.com
Cc: yasu.isimatu@gmail.com
Link: http://lkml.kernel.org/r/20180625031656.12443-3-bhe@redhat.com
[ Rewrote the changelog, fixed style problems in the code. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
