Commit graph

670 commits

Author SHA1 Message Date
Simon Glass
42fd8c19b5 x86: Use unsigned long for address in table generation
We should use unsigned long rather than u32 for addresses. Update this so
that the table-generation code builds correctly on 64-bit machines.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2017-02-06 11:38:46 +08:00
Mario Six
7690be35de lib: tpm: Add command to flush resources
This patch adds a function to the TPM library, which allows U-Boot to
flush resources, e.g. keys, from the TPM.

Signed-off-by: Mario Six <mario.six@gdsys.cc>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Stefan Roese <sr@denx.de>
2017-02-01 09:02:57 +01:00
Patrick Delaunay
1acc008787 disk: convert CONFIG_ISO_PARTITION to Kconfig
Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
Signed-off-by: Patrick Delaunay <patrick.delaunay73@gmail.com>
2017-01-28 08:47:35 -05:00
George McCollister
f1ca1fdebf mkimage: Add support for signing with pkcs11
Add support for signing with the pkcs11 engine. This allows FIT images
to be signed with keys securely stored on a smartcard, hardware security
module, etc without exposing the keys.

Support for other engines can be added in the future by modifying
rsa_engine_get_pub_key() and rsa_engine_get_priv_key() to construct
correct key_id strings.

Signed-off-by: George McCollister <george.mccollister@gmail.com>
2017-01-14 16:47:13 -05:00
Chris Packham
f267e40f96 lib: net_utils: enforce '.' as octet separator in string_to_ip
Ensure '.' is used to separate octets. If another character is seen
reject the string outright and return 0.0.0.0.

Signed-off-by: Chris Packham <judge.packham@gmail.com>
2017-01-14 16:47:11 -05:00
Chris Packham
d921ed9a2a lib: net_utils: make string_to_ip stricter
Previously values greater than 255 were implicitly truncated. Add some
stricter checking to reject addresses with components >255.

With the input "1234192.168.1.1" the old behaviour would truncate the
address to 192.168.1.1. New behaviour rejects the string outright and
returns 0.0.0.0, which for the purposes of IP addresses can be
considered an error.

Signed-off-by: Chris Packham <judge.packham@gmail.com>
2017-01-14 16:47:11 -05:00
Masahiro Yamada
5bc516ed66 delay: collect {m, n, u}delay declarations to include/linux/delay.h
Currently, mdelay() and udelay() are declared in include/common.h,
while ndelay() in include/linux/compat.h.  It would be nice to
collect them into include/linux/delay.h like Linux.

While we are here, fix the ndelay() implementation; I used the
DIV_ROUND_UP() instead of (x)/1000 because it must wait *longer*
than the given period of time.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2017-01-14 16:46:28 -05:00
David Gibson
46743c412d libfdt: Correct fdt handling of overlays without fixups and base trees without symbols
The fdt_overlay_apply() function purports to support the edge cases where
an overlay has no fixups to be applied, or a base tree which has no
symbols (the latter can only work if the former is also true).  However it
gets it wrong in a couple of small ways:

  * In the no fixups case, it doesn't fail immediately, but will attempt
    fdt_for_each_property_offset() giving -FDT_ERR_NOTFOUND as the node
    offset, which will fail.  Instead it should succeed immediately, since
    there's nothing to do.
  * In the case of no symbols, it again doesn't fail immediately.  However
    if there is an actual fixup it will fail with an unexpected error,
    because -FDT_ERR_NOTFOUND is passed to fdt_getprop() when attempting to
    look up the symbols.  We should instead return -FDT_ERR_NOTFOUND
    directly.

Both of these errors lead to the code returning misleading error codes in
failing cases.

[ DTC commit: 7d8ef6e1db9794f72805a0855f4f7f12fadd03d3 ]

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Simon Glass <sjg@chromium.org>
2017-01-13 18:19:45 -07:00
Tom Rini
0b8404332e Merge branch 'master' of git://git.denx.de/u-boot-sunxi 2017-01-10 08:19:21 -05:00
Ladislav Michl
8361af0d30 lib: gitignore *.elf and *.so generated by efi_loader
Signed-off-by: Ladislav Michl <ladis@linux-mips.org>
2017-01-09 10:30:24 -05:00
Andre Przywara
1c853629d9 SPL: tiny-printf: ignore "-" modifier
tiny-printf does not know about the "-" modifier, which aligns numbers.
This is used by some SPL code, but as it's purely cosmetical, we just
ignore this modifier here to avoid changing correct printf strings.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Jagan Teki <jagan@openedev.com>
2017-01-04 16:37:41 +01:00
Andre Przywara
a28e1d9831 SPL: tiny-printf: add "l" modifier
tiny-printf does not know about the "l" modifier so far, which breaks
the crash dump on AArch64, because it uses %lx to print the registers.
Add an easy way of handling longs correctly.

Using a relatively decent compiler (GCC 5.3.0) this does _not_ increase
the code size of tiny-printf.o for 32-bit builds (where long and int
are actually the same), actually it looses three (ARM Thumb2) instructions
from the actual SPL (numbers for orangepi_plus_defconfig):
  text    data     bss     dec     hex filename
   758       0       0     758     2f6 spl/lib/tiny-printf.o	before
 18839     488     232   19559    4c67 spl/u-boot-spl		before
   758       0       0     758     2f6 spl/lib/tiny-printf.o	after
 18833     488     232   19553    4c61 spl/u-boot-spl		after

This adds some substantial amount of code to a 64-bit build, though:
(taken after a later commit, which enables the ARM64 SPL build for sunxi)
  text    data     bss     dec     hex filename
  1542       0       0    1542     606 spl/lib/tiny-printf.o	before
 25830     392     360   26582    67d6 spl/u-boot-spl		before
  1758       0       0    1758     6de spl/lib/tiny-printf.o	after
 26040     392     360   26792    68a8 spl/u-boot-spl		after

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Jagan Teki <jagan@openedev.com>
2017-01-04 16:37:40 +01:00
Nathan Rossi
623f60198b fdt: add memory bank decoding functions for board setup
Add two functions for use by board implementations to decode the memory
banks of the /memory node so as to populate the global data with
ram_size and board info for memory banks.

The fdtdec_setup_memory_size() function decodes the first memory bank
and sets up the gd->ram_size with the size of the memory bank. This
function should be called from the boards dram_init().

The fdtdec_setup_memory_banksize() function decode the memory banks
(up to the CONFIG_NR_DRAM_BANKS) and populates the base address and size
into the gd->bd->bi_dram array of banks. This function should be called
from the boards dram_init_banksize().

Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Michal Simek <monstr@monstr.eu>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
2016-12-20 09:15:28 +01:00
Alexander Graf
95b62b2e28 efi_loader: Allow to compile helloworld.efi w/o bundling it
Today we can compile a self-contained hello world efi test binary that
allows us to quickly verify whether the EFI loader framwork works.

We can use that binary outside of the self-contained test case though,
by providing it to a to-be-tested system via tftp.

This patch separates compilation of the helloworld.efi file from
including it in the u-boot binary for "bootefi hello". It also modifies
the efi_loader test case to enable travis to pick up the compiled file.
Because we're now no longer bloating the resulting u-boot binary, we
can enable compilation always, giving us good travis test coverage.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
2016-11-27 09:53:39 -05:00
Andrew Duda
83dd98e012 image: Combine image_sig_algo with image_sign_info
Remove the need to explicitly add SHA/RSA pairings. Invalid SHA/RSA
pairings will still fail on verify operations when the hash length is
longer than the key length.

Follow the same naming scheme "checksum,crytpo" without explicitly
defining the string.

Indirectly adds support for "sha1,rsa4096" signing/verification.

Signed-off-by: Andrew Duda <aduda@meraki.com>
Signed-off-by: aduda <aduda@meraki.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-11-21 14:07:31 -05:00
Andrew Duda
0c1d74fda7 image: Add crypto_algo struct for RSA info
Cut down on the repetition of algorithm information by defining separate
checksum and crypto structs. image_sig_algos are now simply pairs of
unique checksum and crypto algos.

Signed-off-by: Andrew Duda <aduda@meraki.com>
Signed-off-by: aduda <aduda@meraki.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-11-21 14:07:31 -05:00
Andrew Duda
da29f2991d rsa: Verify RSA padding programatically
Padding verification was done against static SHA/RSA pair arrays which
take up a lot of static memory, are mostly 0xff, and cannot be reused
for additional SHA/RSA pairings. The padding can be easily computed
according to PKCS#1v2.1 as:

  EM = 0x00 || 0x01 || PS || 0x00 || T

where PS is (emLen - tLen - 3) octets of 0xff and T is DER encoding
of the hash.

Store DER prefix in checksum_algo and create rsa_verify_padding
function to handle verification of a message for any SHA/RSA pairing.

Signed-off-by: Andrew Duda <aduda@meraki.com>
Signed-off-by: aduda <aduda@meraki.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-11-21 14:07:30 -05:00
Andrew Duda
5300a4f933 rsa: cosmetic: rename pad_len to key_len
checksum_algo's pad_len field isn't actually used to store the length of
the padding but the total length of the RSA key (msg_len + pad_len)

Signed-off-by: Andrew Duda <aduda@meraki.com>
Signed-off-by: aduda <aduda@meraki.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-11-21 14:07:30 -05:00
Alexander Graf
b7b8410a8f ls2080: Exit dpaa only right before exiting U-Boot
On ls2080 we have a separate network fabric component which we need to
shut down before we enter Linux (or any other OS). Along with that also
comes configuration of the fabric using a description file.

Today we always stop and configure the fabric in the boot script and
(again) exit it on device tree generation. This works ok for the normal
booti case, but with bootefi the payload we're running may still want to
access the network.

So let's add a new fsl_mc command that defers configuration and stopping
the hardware to when we actually exit U-Boot, so that we can still use
the fabric from an EFI payload.

For existing boot scripts, nothing should change with this patch.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: York Sun <york.sun@nxp.com>
[agraf: Fix x86 build]
2016-11-17 14:18:55 +01:00
Simon Glass
58ad86288f x86: Enable EFI loader support
Enable this so that EFI applications (notably grub) can be run under U-Boot
on x86 platforms.

At present the 'hello world' EFI application is not supported for the
qemu-x86_efi_payload64 board. That board builds a payload consisting of a
64-bit header and a 32-bit U-Boot, which is incompatible with the way the
EFI loader builds its EFI application. The following error is obtained:

x86_64-linux-ld.bfd: i386 architecture of input file
   `lib/efi_loader/helloworld.o' is incompatible with i386:x86-64 output

This could be corrected with additional Makefile rules. For now, this
feature is disabled for that board.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
[agraf: drop hello kconfig bits]
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:04 +01:00
Simon Glass
c7ae3dfdcc efi: Add support for a hello world test program
It is useful to have a basic sanity check for EFI loader support. Add a
'bootefi hello' command which loads HelloWord.efi and runs it under U-Boot.

Signed-off-by: Simon Glass <sjg@chromium.org>
[agraf: Fix documentation, add unfulfilled kconfig dep]
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:03 +01:00
Simon Glass
5abd9137d5 x86: Tidy up selection of building the EFI stub
At present we use a CONFIG option in efi.h to determine whether we are
building the EFI stub or not. This means that the same header cannot be
used for EFI_LOADER support. The CONFIG option will be enabled for the
whole build, even when not building the stub.

Use a different define instead, set up just for the files that make up the
stub.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:03 +01:00
Simon Glass
d0d9099365 efi: Correct cache flush alignment
Make sure that the cache flushes correctly by ensuring that the end
address is correctly aligned.

Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:03 +01:00
Emmanuel Vadot
5be8b0a338 efi_loader: console: Correctly report modes
Add support for EFI console modes.
Mode 0 is always 80x25 and present by EFI specification.
Mode 1 is always 80x50 and not mandatory.
Mode 2 and above is freely usable.

If the terminal can handle mode 1, we mark it as supported.
If the terminal size is greater than mode 0 and different than mode 1,
we install it as mode 2.

Modes can be switch with cout_set_mode.

Changes in V5:
 Correctly detect mode before enabling mode 2.

Changes in V4:
 Reset cursor positon on mode switch
 Use local variables in console query code

Changes in V3:
 Valid mode are 0 to EFIMode-1
 Fix style

Changes in V2:
 Add mode switch
 Report only the modes that we support

Signed-off-by: Emmanuel Vadot <manu@bidouilliste.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:02 +01:00
Oleksandr Tymoshenko
d7608aba38 efi: Use device device path type Messaging for network interface node
When adding network interface node use Messaging device path with
subtype MAC Address and device's MAC address as a value instead
of Media Device path type with subtype File Path and path "Net"

Signed-off-by: Oleksandr Tymoshenko <gonzo@bluezbox.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:02 +01:00
Masahiro Yamada
456ca6ba04 efi_loader: fix depends on line of EFI_LOADER
This line is shown as

   depends on (ARM64 ||\302\240ARM) && OF_LIBFDT

on my Emacs.  Use ASCII characters only.

Assuming it is (ARM64 || ARM), remove the redundancy.
Unlike Linux, CONFIG_ARM includes CONFIG_ARM64 in U-Boot.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-11-14 23:24:02 +01:00
Masahiro Yamada
01ae56cfcb libfdt: fix fdt_stringlist_search()
If fdt_getprop() fails, negative error code should be returned.

[ DTC commit: daa75e8fa5942caa8e97931aed3a1ee0b7edd74b ]

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-10-24 08:04:40 -04:00
Masahiro Yamada
7c9786d61f libfdt: fix fdt_stringlist_count()
If fdt_getprop() fails, negative error code should be returned.

[ DTC commit: e28eff5b787adb3f461d1653598818b2f1f25a73 ]

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2016-10-24 08:04:39 -04:00
Maxime Ripard
610db7058f libfdt: Sync overlay with upstream
Now that the overlay code has been merge upstream, update our copy to
what's been merged, since a significant number of issues have been fixed
during the merge process.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Simon Glass <sjg@chromium.org>
2016-10-23 12:08:48 -07:00
Alexander Graf
3c63db9ca9 efi_loader: Rename EFI_RUNTIME_{TEXT, DATA} to __efi_runtime{, _data}
Compiler attributes are more commonly __foo style tags rather than big
upper case eye sores like EFI_RUNTIME_TEXT.

Simon Glass felt quite strongly about this, so this patch converts our
existing defines over to more eye friendly ones.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:54 +02:00
Simon Glass
65e4c0b168 x86: efi: Add EFI loader support for x86
Add the required pieces to support the EFI loader on x86.

Since U-Boot only builds for 32-bit on x86, only a 32-bit EFI application
is supported. If a 64-bit kernel must be booted, U-Boot supports this
directly using FIT (see doc/uImage.FIT/kernel.its). U-Boot can act as a
payload for both 32-bit and 64-bit EFI.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-19 09:01:53 +02:00
Simon Glass
e275458c2f efi: Fix missing EFIAPI specifiers
These are missing in some functions. Add them to keep things consistent.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-19 09:01:53 +02:00
Alexander Graf
6fb580d7b4 smbios: Provide serial number
If the system has a valid "serial#" environment variable set (which boards that
can find it out programatically set automatically), use that as input for the
serial number and UUID fields in the SMBIOS tables.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:53 +02:00
Alexander Graf
aba5e9194b efi_loader: Fix efi_install_configuration_table
So far we were only installing the FDT table and didn't have space
to store any other. Hence nobody realized that our efi table allocation
was broken in that it didn't set the indicator for the number of tables
plus one.

This patch fixes it, allowing code to allocate new efi tables.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:52 +02:00
Alexander Graf
e663b350f1 smbios: Expose in efi_loader as table
We can pass SMBIOS easily as EFI configuration table to an EFI payload. This
patch adds enablement for that case.

While at it, we also enable SMBIOS generation for ARM systems, since they support
EFI_LOADER.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:52 +02:00
Alexander Graf
96476206c5 smbios: Generate type 4 on non-x86 systems
The type 4 table generation code is very x86 centric today. Refactor things
out into the device model cpu class to allow the tables to get generated for
other architectures as well.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2016-10-19 09:01:52 +02:00
Alexander Graf
e824cf3fb5 smbios: Allow compilation on 64bit systems
The SMBIOS generation code passes pointers as u32. That causes the compiler
to warn on casts to pointers. This patch moves all address pointers to
uintptr_t instead.

Technically u32 would be enough for the current SMBIOS2 style tables, but
we may want to extend the code to SMBIOS3 in the future which is 64bit
address capable.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:51 +02:00
Alexander Graf
488bf12d84 efi_loader: Expose efi_install_configuration_table
We want to be able to add configuration table entries from our own code as
well as from EFI payload code. Export the boot service function internally
too, so that we can reuse it.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:51 +02:00
Alexander Graf
4b6dddc294 x86: Move smbios generation into arch independent directory
We will need the SMBIOS generation function on ARM as well going forward,
so let's move it into a non arch specific location.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2016-10-19 09:01:50 +02:00
Alexander Graf
1befb38b86 x86: Move table csum into separate file
We need the checksum function without all the other table functionality
soon, so let's split it out into its own C file.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
2016-10-19 09:01:50 +02:00
Alexander Graf
8f661a5b66 efi_loader: gop: Expose fb when 32bpp
When we're running in 32bpp mode, expose the frame buffer address
to our payloads so that Linux efifb can pick it up.

Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-19 09:01:50 +02:00
Alexander Graf
712cd29874 efi_loader: Allow bouncing for network
So far bounce buffers were only used for disk I/O, but network I/O
may suffer from the same problem.

On platforms that have problems doing DMA on high addresses, let's
also bounce outgoing network packets. Incoming ones always already
get bounced.

This patch fixes EFI PXE boot on ZynqMP for me.

Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-19 09:01:50 +02:00
Alexander Graf
80a4800ee1 efi_loader: Allow boards to implement get_time and reset_system
EFI allows an OS to leverage firmware drivers while the OS is running. In the
generic code we so far had to stub those implementations out, because we would
need board specific knowledge about MMIO setups for it.

However, boards can easily implement those themselves. This patch provides the
framework so that a board can implement its own versions of get_time and
reset_system which would actually do something useful.

While at it we also introduce a simple way for code to reserve MMIO pointers
as runtime available.

Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:08 +02:00
Stefan Brüns
511d0b97ef efi_loader: Do not leak memory when unlinking a mapping
As soon as a mapping is unlinked from the list, there are no further
references to it, so it should be freed. If it not unlinked,
update the start address and length.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:08 +02:00
Stefan Brüns
b6a9517275 efi_loader: Keep memory mapping sorted when splitting an entry
The code assumes sorted mappings in descending address order. When
splitting a mapping, insert the new part next to the current mapping.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00
Stefan Brüns
b61d857b2f efi_loader: Readd freed pages to memory pool
Currently each allocation creates a new mapping. Readding the mapping
as free memory (EFI_CONVENTIONAL_MEMORY) potentially allows to hand out
an existing mapping, thus limiting the number of mapping descriptors in
the memory map.

Mitigates a problem with current (4.8rc7) linux kernels when doing an
efi_get_memory map, resulting in an infinite loop. Space for the memory
map is reserved with allocate_pool (implicitly creating a new mapping) and
filled. If there is insufficient slack space (8 entries) in the map, the
space is freed and a new round is started, with space for one more entry.
As each round increases requirement and allocation by exactly one, there
is never enough slack space. (At least 32 entries are allocated, so as
long as there are less than 24 entries, there is enough slack).
Earlier kernels reserved no slack, and did less allocations, so this
problem was not visible.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00
Stefan Brüns
42417bc84d efi_loader: Track size of pool allocations to allow freeing
We need a functional free_pool implementation, as otherwise each
allocate_pool causes growth of the memory descriptor table.

Different to free_pages, free_pool does not provide the size for the
to be freed allocation, thus we have to track the size ourselves.

As the only EFI requirement for pool allocation is an alignment of
8 bytes, we can keep allocating a range using the page allocator,
reserve the first 8 bytes for our bookkeeping and hand out the
remainder to the caller. This saves us from having to use any
independent data structures for tracking.

To simplify the conversion between pool allocations and the corresponding
page allocation, we create an auxiliary struct efi_pool_allocation.

Given the allocation size free_pool size can handoff freeing the page
range, which was indirectly allocated by a call to allocate_pool,
to free_pages.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00
Stefan Brüns
ead1274b7f efi_loader: Move efi_allocate_pool implementation to efi_memory.c
We currently handle efi_allocate_pool() in our boot time service
file. In the following patch, pool allocation will receive additional
internal semantics that we should preserve inside efi_memory.c instead.

As foundation for those changes, split the function into an externally
facing efi_allocate_pool_ext() for use by payloads and an internal helper
efi_allocate_pool() in efi_memory.c that handles the actual allocation.

While at it, change the magic 0xfff / 12 constants to the more obvious
EFI_PAGE_MASK/SHIFT defines.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00
Robin Randhawa
991d62fa73 efi_loader: Fix crash on 32-bit systems
A type mismatch in the efi_allocate_pool boot service flow causes
hazardous memory scribbling on 32-bit systems.

This is efi_allocate_pool's prototype:

static efi_status_t EFIAPI efi_allocate_pool(int pool_type,
						    unsigned long size,
						    void **buffer);

Internally, it invokes efi_allocate_pages as follows:

efi_allocate_pages(0, pool_type, (size + 0xfff) >> 12,
					    (void*)buffer);

This is efi_allocate_pages' prototype:

efi_status_t efi_allocate_pages(int type, int memory_type,
					unsigned long pages,
					uint64_t *memory);

The problem: efi_allocate_pages does this internally:

    *memory = addr;

This fix in efi_allocate_pool uses a transitional uintptr_t cast to
ensure the correct outcome, irrespective of the system's native word
size.

This was observed when bootefi'ing the EFI instance of FreeBSD's first
stage bootstrap (boot1.efi) on a 32-bit ARM platform (Qemu VExpress +
Cortex-a9).

Signed-off-by: Robin Randhawa <robin.randhawa@arm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00
Stefan Brüns
bdf5c1b360 efi_loader: Fix memory map size check to avoid out-of-bounds access
The current efi_get_memory_map() function overwrites the map_size
property before reading its value. That way the sanity check whether our
memory map fits into the given array always succeeds, potentially
overwriting arbitrary payload memory.

This patch moves the property update write after its sanity check, so
that the check actually verifies the correct value.

So far this has not triggered any known bugs, but we're better off safe
than sorry.

If the buffer is to small, the returned memory_map_size indicates the
required size to the caller.

Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
2016-10-18 09:08:07 +02:00