Skip to content

Instantly share code, notes, and snippets.

@roadrunner2
Last active November 15, 2024 17:51
Show Gist options
  • Save roadrunner2/1289542a748d9a104e7baec6a92f9cd7 to your computer and use it in GitHub Desktop.
Save roadrunner2/1289542a748d9a104e7baec6a92f9cd7 to your computer and use it in GitHub Desktop.
Linux on MacBook Pro Late 2016 and Mid 2017 (with Touchbar)

Introduction

This is about documenting getting Linux running on the late 2016 and mid 2017 MPB's; the focus is mostly on the MacBookPro13,3 and MacBookPro14,3 (15inch models), but I try to make it relevant and provide information for MacBookPro13,1, MacBookPro13,2, MacBookPro14,1, and MacBookPro14,2 (13inch models) too. I'm currently using Fedora 27, but most the things should be valid for other recent distros even if the details differ. The kernel version is 4.14.x (after latest update).

The state of linux on the MBP (with particular focus on MacBookPro13,2) is also being tracked on https://github.com/Dunedan/mbp-2016-linux . And for Ubuntu users there are a couple tutorials (here and here) focused on that distro and the MacBook.

Note: For those who have followed these instructions ealier, and in particular for those who have had problems with the custom DSDT, modifying the DSDT is not necessary anymore - see the updated instructions below and make sure to update your clone of the roadrunner2/macbook12-spi-driver repo to get the latest drivers.

Summary Of Current State

What works

  • Booting (i.e Grub etc)
  • Recognizes disk on all models (older kernels may need patch for some models, though)
  • Keyboard, touchpad, and basic touchbar functionality
  • HiDPI detection
  • Accelerated video
  • Screen brightness control
  • Keyboard backlight
  • USB
  • Sensors (install lm_sensors package)
  • Camera
  • Bluetooth (older kernels need patches)
  • WiFi on MBP13,1 and MBP14,1
  • Thunderbolt
  • DisplayPort

What doesn't work

  • WiFi on ,2 and ,3 models (though some folks have had success with some of the workarounds)
  • Suspend/Resume (works partially on 13,3 models)
  • Audio (two cards show up, and intel driver is loaded, but no sound)

Details

Partitioning

If you want to keep your MacOS installation (generally a good idea if you can afford the disk space, because that's the only way to get/install firmware updates), then first boot into MacOS and resize the partition there, creating a new partition for the Linux installation. If you also want to have a Windows partition, see this comment below.

Warning: If you're not going to keep MacOS, either back up the EFI System Partition (and restore its contents to the new ESP after installation) or leave it intact (i.e. don't do a full disk install, but just use the space after the ESP). This partition (it's the first one) contains drivers/firmware/etc needed by Apple's EFI loader during boot, in particular to initialize the Touchbar.

Initial Installation

Since the internal keyboard and touchpad won't work until you have built and loaded the drivers, you'll need to plug in an external USB keyboard to do the initial setup and installation.

Booting

If you're booting a 4.11 or later kernel, no special params or patches are needed.

If you're booting a kernel < 4.11 and have a MacBookPro13,1, MacBookPro13,2, MacBookPro14,1 or MacBookPro14,2 (13inch models), which have the Apple NVMe controller, you'll need the kernel-nvme-controller.patch from this gist in order for the disk to be correctly recognized (MacBookPro13,3 uses a Samsung NVMe controller which is automatically detected correctly). Alternatively, instead of patching you can also do the following (for distros using something other than dracut to create the initrd you'll need to adjust the 2nd and 3rd lines appropriately):

echo 'install nvme /sbin/modprobe --ignore-install nvme $CMDLINE_OPTS; echo 106b 2003 > /sys/bus/pci/drivers/nvme/new_id' | sudo tee /etc/modprobe.d/nvme.conf
echo 'force_drivers+="nvme"' | sudo tee /etc/dracut.conf.d/disk.conf
sudo dracut --force --kver <kernel-version>

If you're booting a kernel < 4.10 then you'll need the following kernel param to boot properly: intremap=nosid. E.g.

sudo sed -i 's/\(GRUB_CMDLINE_LINUX=.*\)"/\1 intremap=nosid"/' /etc/default/grub
sudo grub2-mkconfig -o /boot/efi/EFI/fedora/grub.cfg

Lastly, if you are booting a live CD or similar with a kernel < 4.9 then you will also need to add the nomodeset kernel parameter to your kernel line; you will then not have proper HiDPI detection or accelerated graphics.

Keyboard/Touchpad/Touchbar

For this we need the drivers from https://github.com/roadrunner2/macbook12-spi-driver.git (a clone of https://github.com/cb22/macbook12-spi-driver which includes a preliminary touchbar driver and keyboard fixes). The following commands set this up.

First some extra packages:

sudo dnf install git kernel-devel dkms

Next we need to prepare for the modules to be included in the ramdisk (so they are loaded early during boot):

cat <<EOF | sudo tee /etc/dracut.conf.d/keyboard.conf
# load all drivers needed for the keyboard+touchpad
add_drivers+="applespi intel_lpss_pci spi_pxa2xx_platform apple-ib-tb"
EOF

On distros using mkinitramfs instead of dracut you'll want to do the following instead:

cat <<EOF | sudo tee -a /etc/initramfs-tools/modules
# drivers for keyboard+touchpad
applespi
apple-ib-tb
intel_lpss_pci
spi_pxa2xx_platform
EOF

Now get and build the drivers:

git clone https://github.com/roadrunner2/macbook12-spi-driver.git
pushd macbook12-spi-driver
git checkout touchbar-driver-hid-driver
sudo ln -s `pwd` /usr/src/applespi-0.1
sudo dkms install applespi/0.1
popd

Next we need to set the proper dpi for the touchpad and adjust the sensitivity (download the 61-evdev-local.hwdb, 61-libinput-local.hwdb, and local-overrides.quirks from this gist - the 61-evdev-local.hwdb is only needed for udev < 242, the 61-libinput-local.hwdb is needed for libinput versions < 1.12, the local-overrides.quirks for libinput versions >= 1.12):

# if 'udevadm --version' less than 242:
sudo cp ...the-downloaded-61-evdev-local.hwdb... /etc/udev/hwdb.d/61-evdev-local.hwdb
# if 'libinput --version' less than 1.12:
sudo cp ...the-downloaded-61-libinput-local.hwdb... /etc/udev/hwdb.d/61-libinput-local.hwdb
# if 'libinput --version' 1.12 or later:
sudo cp ...the-downloaded-local-overrides.quirks... /etc/libinput/local-overrides.quirks
sudo systemd-hwdb update

You can test the drivers by loading them and their dependencies:

sudo modprobe intel_lpss_pci spi_pxa2xx_platform applespi apple-ib-tb

Finally, reboot to make sure it all works correctly:

sudo reboot

Screen Brightness Control

Screen brightness control works out of the box on MacBookPro13,1 and MacBookPro13,2 (all kernels), and MacBookPro13,3 with recent kernels, but requires a kernel patch on MacBookPro13,3 with older kernels (see also Dunedan/mbp-2016-linux#2). Specifically, if you have any of these kernels you need to patch: < 4.14, 4.14 - 4.14.21, 4.15 - v4.15.5 (i.e. the issue was fixed in 4.14.21, 4.15.5, and 4.16). The following will create and install the patched apple-gmux:

mkdir apple-gmux
pushd apple-gmux

curl -o apple-gmux.patch 'https://bugzilla.kernel.org/attachment.cgi?id=192601'
curl -o apple-gmux.c 'https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/plain/drivers/platform/x86/apple-gmux.c?id=refs/tags/v4.9.11'

patch < apple-gmux.patch

echo -e '
obj-m += apple-gmux.o

all:
\tmake -C /lib/modules/`uname -r`/build M=`pwd` modules
' > Makefile
make

mod=$(ls /lib/modules/`uname -r`/kernel/drivers/platform/x86/apple-gmux.ko*)
sudo mv $mod{,.orig}
sudo cp apple-gmux.ko /lib/modules/`uname -r`/kernel/drivers/platform/x86/
sudo depmod

popd

sudo reboot

Other

The touchpad defaults to using the bottom-left corner for right-clicks - to get 2-finger right click, install the Gnome tweak tool and change it in there.

WiFi

WiFi works fine on MBP13,1 and MBP14,1. But the other models use a different chipset, and while on those the brcmfmac driver is automatically loaded, there are a number of issues with it, making it for all practical purposes unusable:

  • it only does 2.4GHz - no 5GHz channels are visible
  • it has an extremely low sensitivity - you must be within a few feet of the base station, and even at 5 feet distance it shows a weak signal.
  • it stops working after 10 or 15 or so minutes; turning WiFi off, waiting a several minutes, and then turning it back on generally gets it working again. Maybe a thermal issue?

Bug report: https://bugzilla.kernel.org/show_bug.cgi?id=193121

In the mean time some folks have that one or both of the following hacks make the WiFi work well enough them (personally, while they do improve the situation, I have not found them to be sufficient enough for actual work, i.e. I still see many packet drops and connection failures - YMMV):

  • reduce the transmit power: sudo iwconfig wlp3s0 txpower 10
  • edit the firmware blob (/lib/firmware/brcm/brcmfmac43602-pcie.bin) and modify the regrev and ccode values (see the above bugreport for details)

Display

The amdgpu driver works well and is automatically loaded on MacBookPro13,3. On the 13 inch models the use of the intel Xorg driver may need to be forced (see first comment below).

Using the iGPU (*,3 models only)

The *,3 models have both an iGPU (intel) as well as a dGPU (amd). By default the dGPU is used. While this works well, it does use a lot more power than the iGPU. The following will set things up so the machine uses the iGPU upon boot (but see the caveats below):

mkdir gpu-switch
cd gpu-switch

# Enable the iGPU at boot
curl -O -L https://github.com/0xbb/apple_set_os.efi/releases/download/v1/apple_set_os.efi

sudo mkdir /boot/efi/EFI/custom
sudo cp apple_set_os.efi /boot/efi/EFI/custom

config_dir=$(dirname $(sudo find /boot -name grub.cfg))
cat <<"EOF" | sudo tee "$config_dir"/custom.cfg > /dev/null
insmod chain
search --no-floppy --set=cuspart --file /EFI/custom/apple_set_os.efi
chainloader (${cuspart})/EFI/custom/apple_set_os.efi
boot
EOF

# Switch the default gpu on boot to iGPU
curl -O -L https://raw.githubusercontent.com/0xbb/gpu-switch/master/gpu-switch
chmod +x gpu-switch

./gpu-switch -i

# Service to disable the dGPU
cat <<EOF | sudo tee /etc/systemd/system/disable-dgpu.service > /dev/null
[Unit]
Description=Disable the dGPU

[Service]
Type=oneshot
ExecStart=/bin/sh -c 'echo OFF > /sys/kernel/debug/vgaswitcheroo/switch'
ExecStop=/bin/sh -c 'echo ON > /sys/kernel/debug/vgaswitcheroo/switch'
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target
EOF

sudo chmod 644 /etc/systemd/system/disable-dgpu.service

cd -

Once you then disable the dGPU this will save save you about 5W - 6W, or ⅓ of the power. However, there are several caveats to aware of:

  • The dGPU needs to be manually disabled after each boot and first login with sudo systemctl start disable-dgpu - trying to have it run automatically at boot via systemctl enable disable-gpu leads to hangs.
  • In order to use an external display (via Thunderbolt/DisplayPort) the dGPU needs to be enabled again (the iGPU is not wired up for that): sudo systemctl disable disable-dgpu
  • The amd driver does not handle disabling and re-enabling the of the dGPU very well. In particular system shutdown may hang or just take several minutes, and you may see various errors in the kernel logs when disabling and re-enabling.
  • If you boot into MacOS, then when booting back into Linux you'll need to change the default gpu back to the iGPU again: boot, run gpu-switch/gpu-switch -i, and reboot.

Camera

MacBookPro13,3/14,3: works out of the box on kernels 4.13 and later; on earlier kernels you need the following:

echo "options uvcvideo quirks=0x100" > /etc/modprobe.d/uvcvideo.conf

For MacBookPro[13,14],[12] you need the bcwc_pcie driver (mainline branch) - see also Dunedan/mbp-2016-linux#15.

Bluetooth

As of kernel 4.16 bluetooth works out of the box; older kernels need patches - see Dunedan/mbp-2016-linux#29 (comment) and following discussion for details. But in short you'll need to:

  • Ensure your kernel is configured with CONFIG_BT_HCIUART_BCM=y
  • apply the patches from hci_bcm-4.13, hci_bcm-4.14, or hci_bcm-4.15, depending on whether you have a 4.13 or earlier, 4.14, or 4.15 kernel.
  • build and reboot
  • on 4.14 and earlier apply the service patch from the above comment and start the service as described there (not necessary on 4.15 and later).

Note that as of 4.16 there are still issues on MacBookPro13,1 and MacBookPro14,1 - see the above bug for details on what additional patches are needed.

# MacBook8,1 (2015), MacBook9,1 (2016), MacBook10,1 (2017)
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBook8,1:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBook9,1:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBook10,1:*
EVDEV_ABS_00=::95
EVDEV_ABS_01=::90
EVDEV_ABS_35=::95
EVDEV_ABS_36=::90
# MacBookPro13,* (Late 2016), MacBookPro14,* (Mid 2017)
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro13,1:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro13,2:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro14,1:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro14,2:*
EVDEV_ABS_00=::96
EVDEV_ABS_01=::94
EVDEV_ABS_35=::96
EVDEV_ABS_36=::94
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro13,3:*
evdev:name:Apple SPI Touchpad:dmi:*:svnAppleInc.:pnMacBookPro14,3:*
EVDEV_ABS_00=::96
EVDEV_ABS_01=::95
EVDEV_ABS_35=::96
EVDEV_ABS_36=::95
libinput:name:*Apple SPI Touchpad*:dmi:*
LIBINPUT_MODEL_APPLE_TOUCHPAD=1
LIBINPUT_ATTR_TOUCH_SIZE_RANGE=200:150
LIBINPUT_ATTR_PALM_SIZE_THRESHOLD=1200
libinput:name:*Apple SPI Keyboard*:dmi:*
LIBINPUT_ATTR_KEYBOARD_INTEGRATION=internal
--- a/drivers/nvme/host/pci.c 2017-01-22 21:49:03.710949755 -0800
+++ b/drivers/nvme/host/pci.c 2017-01-22 21:49:23.490761505 -0800
@@ -2121,6 +2121,7 @@
.driver_data = NVME_QUIRK_DELAY_BEFORE_CHK_RDY, },
{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_EXPRESS, 0xffffff) },
{ PCI_DEVICE(PCI_VENDOR_ID_APPLE, 0x2001) },
+ { PCI_DEVICE(PCI_VENDOR_ID_APPLE, 0x2003) },
{ 0, }
};
MODULE_DEVICE_TABLE(pci, nvme_id_table);
[MacBook(Pro) SPI Touchpads]
MatchName=*Apple SPI Touchpad*
ModelAppleTouchpad=1
AttrTouchSizeRange=200:150
AttrPalmSizeThreshold=1100
[MacBook(Pro) SPI Keyboards]
MatchName=*Apple SPI Keyboard*
AttrKeyboardIntegration=internal
[MacBookPro Touchbar]
MatchBus=usb
MatchVendor=0x05AC
MatchProduct=0x8600
AttrKeyboardIntegration=internal
@Bandito11
Copy link

Bandito11 commented Aug 28, 2024

I haven't been able to install the touchbar on kernel 6.10.6-200.fc40.x86_64.

@Drayux
Copy link

Drayux commented Aug 28, 2024

I haven't been able to install the touchbar on kernel 6.10.6-200.fc40.x86_64.

Ah you're not alone! It looks like the API for platform modules changed a bit, so the driver won't be able to build. Shouldn't be too hard of a fix, but I've just started a new job so I haven't found the time to fork the module and do so myself.

I'll look to mention it here when I get the chance to work on it--if no one beats me to it if course!

@l1k
Copy link

l1k commented Aug 29, 2024

There's an effort ongoing to upstream touchbar support into the mainline kernel, so maybe try those patches and report on the mailing list if you come across issues?

https://lore.kernel.org/all/[email protected]/

@iamteedoh
Copy link

I haven't been able to install the touchbar on kernel 6.10.6-200.fc40.x86_64.

Ah you're not alone! It looks like the API for platform modules changed a bit, so the driver won't be able to build. Shouldn't be too hard of a fix, but I've just started a new job so I haven't found the time to fork the module and do so myself.

I'll look to mention it here when I get the chance to work on it--if no one beats me to it if course!

@Drayux - same here! I'm at the build step:

sudo dkms install applespi/0.1
[sudo] password for user:
Sign command: /lib/modules/6.10.9-200.fc40.x86_64/build/scripts/sign-file
Signing key: /var/lib/dkms/mok.key
Public certificate (MOK): /var/lib/dkms/mok.pub
Deprecated feature: REMAKE_INITRD (/var/lib/dkms/applespi/0.1/source/dkms.conf)

Building module:
Cleaning build area...
Building module(s)...(bad exit status: 2)
Failed command:
make -j12 KERNELRELEASE=6.10.9-200.fc40.x86_64
Error! Bad return status for module build on kernel: 6.10.9-200.fc40.x86_64 (x86_64)
Consult /var/lib/dkms/applespi/0.1/build/make.log for more information.

Checking the logs, seems like it's blowing a few chunks but I'm not sure how to handle these yet:

cat /var/lib/dkms/applespi/0.1/build/make.log
DKMS make.log for applespi-0.1 for kernel 6.10.9-200.fc40.x86_64 (x86_64)
Fri Sep 13 11:35:37 AM EDT 2024
make -C /lib/modules/6.10.9-200.fc40.x86_64/build M=/var/lib/dkms/applespi/0.1/build modules
make[1]: Entering directory '/usr/src/kernels/6.10.9-200.fc40.x86_64'
  CC [M]  /var/lib/dkms/applespi/0.1/build/applespi.o
  CC [M]  /var/lib/dkms/applespi/0.1/build/apple-ibridge.o
  CC [M]  /var/lib/dkms/applespi/0.1/build/apple-ib-tb.o
  CC [M]  /var/lib/dkms/applespi/0.1/build/apple-ib-als.o
/var/lib/dkms/applespi/0.1/build/apple-ib-als.c: In function ‘appleals_config_iio’:
/var/lib/dkms/applespi/0.1/build/apple-ib-als.c:463:36: error: passing argument 1 of ‘iio_device_alloc’ makes pointer from integer without a cast [-Wint-conversion]
  463 |         iio_dev = iio_device_alloc(sizeof(als_dev));
      |                                    ^~~~~~~~~~~~~~~
      |                                    |
      |                                    long unsigned int
In file included from ./include/linux/iio/buffer.h:10,
                 from /var/lib/dkms/applespi/0.1/build/apple-ib-als.c:32:
./include/linux/iio/iio.h:761:49: note: expected ‘struct device *’ but argument is of type ‘long unsigned int’
  761 | struct iio_dev *iio_device_alloc(struct device *parent, int sizeof_priv);
      |                                  ~~~~~~~~~~~~~~~^~~~~~
/var/lib/dkms/applespi/0.1/build/apple-ib-als.c:463:19: error: too few arguments to function ‘iio_device_alloc’
  463 |         iio_dev = iio_device_alloc(sizeof(als_dev));
      |                   ^~~~~~~~~~~~~~~~
./include/linux/iio/iio.h:761:17: note: declared here
  761 | struct iio_dev *iio_device_alloc(struct device *parent, int sizeof_priv);
      |                 ^~~~~~~~~~~~~~~~
In file included from /var/lib/dkms/applespi/0.1/build/apple-ib-als.c:36:
/var/lib/dkms/applespi/0.1/build/apple-ib-als.c:485:72: error: ‘struct iio_dev’ has no member named ‘id’
  485 |         iio_trig = iio_trigger_alloc("%s-dev%d", iio_dev->name, iio_dev->id);
      |                                                                        ^~
./include/linux/iio/trigger.h:160:61: note: in definition of macro ‘iio_trigger_alloc’
  160 |         __iio_trigger_alloc((parent), THIS_MODULE, (fmt), ##__VA_ARGS__)
      |                                                             ^~~~~~~~~~~
./include/linux/iio/trigger.h:160:29: error: passing argument 1 of ‘__iio_trigger_alloc’ from incompatible pointer type [-Wincompatible-pointer-types]
  160 |         __iio_trigger_alloc((parent), THIS_MODULE, (fmt), ##__VA_ARGS__)
      |                             ^~~~~~~~
      |                             |
      |                             char *
/var/lib/dkms/applespi/0.1/build/apple-ib-als.c:485:20: note: in expansion of macro ‘iio_trigger_alloc’
  485 |         iio_trig = iio_trigger_alloc("%s-dev%d", iio_dev->name, iio_dev->id);
      |                    ^~~~~~~~~~~~~~~~~
./include/linux/iio/trigger.h:163:56: note: expected ‘struct device *’ but argument is of type ‘char *’
  163 | struct iio_trigger *__iio_trigger_alloc(struct device *parent,
      |                                         ~~~~~~~~~~~~~~~^~~~~~
/var/lib/dkms/applespi/0.1/build/apple-ib-tb.c: In function ‘appletb_fill_report_info’:
/var/lib/dkms/applespi/0.1/build/apple-ib-tb.c:949:9: warning: enumeration value ‘HID_REPORT_TYPES’ not handled in switch [-Wswitch]
  949 |         switch (field->report->type) {
      |         ^~~~~~
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:192:6: warning: no previous prototype for ‘appleib_detach_and_free_hid_driver’ [-Wmissing-prototypes]
  192 | void appleib_detach_and_free_hid_driver(struct appleib_device *ib_dev,
      |      ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
make[3]: *** [scripts/Makefile.build:244: /var/lib/dkms/applespi/0.1/build/apple-ib-als.o] Error 1
make[3]: *** Waiting for unfinished jobs....
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:901:10: error: ‘struct acpi_driver’ has no member named ‘owner’
  901 |         .owner          = THIS_MODULE,
      |          ^~~~~
In file included from ./arch/x86/include/asm/mem_encrypt.h:15,
                 from ./include/linux/mem_encrypt.h:17,
                 from ./arch/x86/include/asm/processor-flags.h:6,
                 from ./arch/x86/include/asm/irqflags.h:5,
                 from ./include/linux/irqflags.h:18,
                 from ./include/linux/spinlock.h:59,
                 from ./include/linux/mmzone.h:8,
                 from ./include/linux/gfp.h:7,
                 from ./include/linux/slab.h:16,
                 from ./include/linux/resource_ext.h:11,
                 from ./include/linux/acpi.h:13,
                 from /var/lib/dkms/applespi/0.1/build/apple-ibridge.c:45:
./include/linux/init.h:180:21: error: initialization of ‘const struct acpi_device_id *’ from incompatible pointer type ‘struct module *’ [-Wincompatible-pointer-types]
  180 | #define THIS_MODULE (&__this_module)
      |                     ^
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:901:27: note: in expansion of macro ‘THIS_MODULE’
  901 |         .owner          = THIS_MODULE,
      |                           ^~~~~~~~~~~
./include/linux/init.h:180:21: note: (near initialization for ‘appleib_driver.ids’)
  180 | #define THIS_MODULE (&__this_module)
      |                     ^
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:901:27: note: in expansion of macro ‘THIS_MODULE’
  901 |         .owner          = THIS_MODULE,
      |                           ^~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:905:35: error: initialization of ‘void (*)(struct acpi_device *)’ from incompatible pointer type ‘int (*)(struct acpi_device *)’ [-Wincompatible-pointer-types]
  905 |                 .remove         = appleib_remove,
      |                                   ^~~~~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/apple-ibridge.c:905:35: note: (near initialization for ‘appleib_driver.ops.remove’)
make[3]: *** [scripts/Makefile.build:244: /var/lib/dkms/applespi/0.1/build/apple-ibridge.o] Error 1
/var/lib/dkms/applespi/0.1/build/applespi.c: In function ‘applespi_setup_read_txfrs’:
/var/lib/dkms/applespi/0.1/build/applespi.c:590:13: error: ‘struct spi_transfer’ has no member named ‘delay_usecs’
  590 |         dl_t->delay_usecs = applespi->spi_settings.spi_cs_delay;
      |             ^~
/var/lib/dkms/applespi/0.1/build/applespi.c: In function ‘applespi_setup_write_txfrs’:
/var/lib/dkms/applespi/0.1/build/applespi.c:619:13: error: ‘struct spi_transfer’ has no member named ‘delay_usecs’
  619 |         wt_t->delay_usecs = SPI_RW_CHG_DELAY_US;
      |             ^~
/var/lib/dkms/applespi/0.1/build/applespi.c:622:13: error: ‘struct spi_transfer’ has no member named ‘delay_usecs’
  622 |         dl_t->delay_usecs = applespi->spi_settings.spi_cs_delay;
      |             ^~
/var/lib/dkms/applespi/0.1/build/applespi.c:626:13: error: ‘struct spi_transfer’ has no member named ‘delay_usecs’
  626 |         wr_t->delay_usecs = SPI_RW_CHG_DELAY_US;
      |             ^~
In file included from ./include/linux/percpu.h:5,
                 from ./arch/x86/include/asm/msr.h:15,
                 from ./arch/x86/include/asm/tsc.h:10,
                 from ./arch/x86/include/asm/timex.h:6,
                 from ./include/linux/timex.h:67,
                 from ./include/linux/time32.h:13,
                 from ./include/linux/time.h:60,
                 from ./include/linux/jiffies.h:10,
                 from ./include/linux/ktime.h:25,
                 from ./include/linux/timer.h:6,
                 from ./include/linux/workqueue.h:9,
                 from ./include/linux/srcu.h:21,
                 from ./include/linux/notifier.h:16,
                 from ./arch/x86/include/asm/uprobes.h:13,
                 from ./include/linux/uprobes.h:49,
                 from ./include/linux/mm_types.h:16,
                 from ./include/linux/mmzone.h:22,
                 from ./include/linux/gfp.h:7,
                 from ./include/linux/slab.h:16,
                 from ./include/linux/resource_ext.h:11,
                 from ./include/linux/acpi.h:13,
                 from /var/lib/dkms/applespi/0.1/build/applespi.c:44:
/var/lib/dkms/applespi/0.1/build/applespi.c: In function ‘applespi_get_saved_bl_level’:
/var/lib/dkms/applespi/0.1/build/applespi.c:1794:38: error: invalid application of ‘sizeof’ to incomplete type ‘struct efivar_entry’
 1794 |         efivar_entry = kmalloc(sizeof(*efivar_entry), GFP_KERNEL);
      |                                      ^
./include/linux/alloc_tag.h:206:16: note: in definition of macro ‘alloc_hooks_tag’
  206 |         typeof(_do_alloc) _res = _do_alloc;                             \
      |                ^~~~~~~~~
./include/linux/slab.h:666:49: note: in expansion of macro ‘alloc_hooks’
  666 | #define kmalloc(...)                            alloc_hooks(kmalloc_noprof(__VA_ARGS__))
      |                                                 ^~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1794:24: note: in expansion of macro ‘kmalloc’
 1794 |         efivar_entry = kmalloc(sizeof(*efivar_entry), GFP_KERNEL);
      |                        ^~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1794:38: error: invalid application of ‘sizeof’ to incomplete type ‘struct efivar_entry’
 1794 |         efivar_entry = kmalloc(sizeof(*efivar_entry), GFP_KERNEL);
      |                                      ^
./include/linux/alloc_tag.h:206:34: note: in definition of macro ‘alloc_hooks_tag’
  206 |         typeof(_do_alloc) _res = _do_alloc;                             \
      |                                  ^~~~~~~~~
./include/linux/slab.h:666:49: note: in expansion of macro ‘alloc_hooks’
  666 | #define kmalloc(...)                            alloc_hooks(kmalloc_noprof(__VA_ARGS__))
      |                                                 ^~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1794:24: note: in expansion of macro ‘kmalloc’
 1794 |         efivar_entry = kmalloc(sizeof(*efivar_entry), GFP_KERNEL);
      |                        ^~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1794:22: error: assignment to ‘struct efivar_entry *’ from ‘int’ makes pointer from integer without a cast [-Wint-conversion]
 1794 |         efivar_entry = kmalloc(sizeof(*efivar_entry), GFP_KERNEL);
      |                      ^
In file included from ./include/linux/string.h:374,
                 from ./include/linux/bitmap.h:13,
                 from ./include/linux/cpumask.h:13,
                 from ./arch/x86/include/asm/paravirt.h:21,
                 from ./arch/x86/include/asm/irqflags.h:60,
                 from ./include/linux/irqflags.h:18,
                 from ./include/linux/spinlock.h:59,
                 from ./include/linux/mmzone.h:8:
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:28: error: invalid use of undefined type ‘struct efivar_entry’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |                            ^~
./include/linux/fortify-string.h:626:34: note: in definition of macro ‘__fortify_memcpy_chk’
  626 |         const size_t __p_size = (p_size);                               \
      |                                  ^~~~~~
./include/linux/fortify-string.h:683:17: note: in expansion of macro ‘__struct_size’
  683 |                 __struct_size(p), __struct_size(q),                     \
      |                 ^~~~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:9: note: in expansion of macro ‘memcpy’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |         ^~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:28: error: invalid use of undefined type ‘struct efivar_entry’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |                            ^~
./include/linux/fortify-string.h:628:40: note: in definition of macro ‘__fortify_memcpy_chk’
  628 |         const size_t __p_size_field = (p_size_field);                   \
      |                                        ^~~~~~~~~~~~
./include/linux/fortify-string.h:684:17: note: in expansion of macro ‘__member_size’
  684 |                 __member_size(p), __member_size(q),                     \
      |                 ^~~~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:9: note: in expansion of macro ‘memcpy’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |         ^~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:28: error: invalid use of undefined type ‘struct efivar_entry’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |                            ^~
./include/linux/fortify-string.h:637:27: note: in definition of macro ‘__fortify_memcpy_chk’
  637 |         __underlying_##op(p, q, __fortify_size);                        \
      |                           ^
/var/lib/dkms/applespi/0.1/build/applespi.c:1798:9: note: in expansion of macro ‘memcpy’
 1798 |         memcpy(efivar_entry->var.VariableName, EFI_BL_LEVEL_NAME,
      |         ^~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:1800:21: error: invalid use of undefined type ‘struct efivar_entry’
 1800 |         efivar_entry->var.VendorGuid = EFI_BL_LEVEL_GUID;
      |                     ^~
/var/lib/dkms/applespi/0.1/build/applespi.c:1803:15: error: implicit declaration of function ‘efivar_entry_get’; did you mean ‘efi_mokvar_entry_next’? [-Wimplicit-function-declaration]
 1803 |         sts = efivar_entry_get(efivar_entry, NULL, &efi_data_len, &efi_data);
      |               ^~~~~~~~~~~~~~~~
      |               efi_mokvar_entry_next
/var/lib/dkms/applespi/0.1/build/applespi.c: In function ‘applespi_save_bl_level’:
/var/lib/dkms/applespi/0.1/build/applespi.c:1830:15: error: implicit declaration of function ‘efivar_entry_set_safe’ [-Wimplicit-function-declaration]
 1830 |         sts = efivar_entry_set_safe((efi_char16_t *)EFI_BL_LEVEL_NAME, efi_guid,
      |               ^~~~~~~~~~~~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c: At top level:
/var/lib/dkms/applespi/0.1/build/applespi.c:2228:27: error: initialization of ‘void (*)(struct spi_device *)’ from incompatible pointer type ‘int (*)(struct spi_device *)’ [-Wincompatible-pointer-types]
 2228 |         .remove         = applespi_remove,
      |                           ^~~~~~~~~~~~~~~
/var/lib/dkms/applespi/0.1/build/applespi.c:2228:27: note: (near initialization for ‘applespi_driver.remove’)
make[3]: *** [scripts/Makefile.build:244: /var/lib/dkms/applespi/0.1/build/applespi.o] Error 1
make[2]: *** [/usr/src/kernels/6.10.9-200.fc40.x86_64/Makefile:1946: /var/lib/dkms/applespi/0.1/build] Error 2
make[1]: *** [Makefile:252: __sub-make] Error 2
make[1]: Leaving directory '/usr/src/kernels/6.10.9-200.fc40.x86_64'
make: *** [Makefile:16: all] Error 2

Thanks in advance!

@stefand
Copy link

stefand commented Nov 15, 2024

With Linux 6.11 I am running into a new issue: The brcmfmac is crashing on load:

[   10.153130] BUG: unable to handle page fault for address: 00007f033c669379
[   10.154402] #PF: supervisor read access in kernel mode
[   10.155665] #PF: error_code(0x0001) - permissions violation
[   10.156913] PGD 8000000101341067 P4D 8000000101341067 PUD 101340067 PMD 1013bb067 PTE 800000010aee9025
[   10.157993] Oops: Oops: 0001 [#1] SMP PTI
[   10.157996] CPU: 4 UID: 0 PID: 827 Comm: (udev-worker) Not tainted 6.11.8-gentoo #1
[   10.158000] Hardware name: Apple Inc. MacBookPro14,3/Mac-551B86E5744E2388, BIOS 529.140.2.0.0 06/23/2024
[   10.158001] RIP: 0010:strlen+0x4/0x30
[   10.158006] Code: f7 75 ec 31 c0 c3 cc cc cc cc 48 89 f8 c3 cc cc cc cc 0f 1f 40 00 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 f3 0f 1e fa <80> 3f 00 74 14 48 89 f8 48 83 c0 01 80 38 00 75 f7 48 29 f8 c3 cc
[   10.158008] RSP: 0018:ffffb4aac0683ad8 EFLAGS: 00010202
[   10.158011] RAX: 00000000ffffffea RBX: 00007f033c669379 RCX: 0000000000000001
[   10.158013] RDX: 0000000000000cc0 RSI: 00007f033c669379 RDI: 00007f033c669379
[   10.158014] RBP: 00000000ffffffea R08: 0000000000000000 R09: 00000000c0ba916a
[   10.169779] R10: ffffffffffffffff R11: ffffffffb61ea260 R12: ffff91f7815b50c8
[   10.169781] R13: 0000000000000cc0 R14: ffff91fafefffe30 R15: ffffb4aac0683b30
[   10.169803] FS:  00007f033ccbe8c0(0000) GS:ffff91faeed00000(0000) knlGS:0000000000000000
[   10.169805] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   10.169806] CR2: 00007f033c669379 CR3: 0000000107b1e004 CR4: 00000000003706f0
[   10.169808] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   10.169809] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   10.169810] Call Trace:
[   10.169812]  <TASK>
[   10.181768]  ? __die+0x23/0x70
[   10.181794]  ? page_fault_oops+0x149/0x4c0
[   10.181797]  ? raw_spin_rq_lock_nested+0xe/0x20
[   10.181799]  ? sched_balance_newidle+0x22b/0x3c0
[   10.181803]  ? update_load_avg+0x78/0x770
[   10.181806]  ? exc_page_fault+0x6f/0x150
[   10.181809]  ? asm_exc_page_fault+0x26/0x30
[   10.190817]  ? __pfx_pci_conf1_write+0x10/0x10
[   10.190822]  ? strlen+0x4/0x30
[   10.190825]  devm_kstrdup+0x25/0x70
[   10.190831]  brcmf_of_probe+0x273/0x350 [brcmfmac]

I could work around the problem with a simple change in drivers/net/wireless/broadcom/brcm80211/brcmfmac/of.c:

@@ -98,7 +98,7 @@
        root = of_find_node_by_path("/");
        if (root && err) {
                char *board_type;
-               const char *tmp;
+               const char *tmp = NULL;
 
                of_property_read_string_index(root, "compatible", 0, &tmp);

I'll try to find the time to do a bisect in the next few days. This file hasn't changed lately, so it's probably not as easy as sloppy error handling in new code. Still, if this is what it looks like (uninitialized pointer being read due to incorrect error handling) then I'd be shocked it made it into the upstream kernel.

Judging from the surrounding code, you might be able to avoid the bug by disabling CONFIG_BRCMFMAC_SDIO. I haven't tested this though.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment