The PCI Vendor and Device ID locations located in firmware were mistakenly
swapped in the bcm5719-fw repository. As a result, the code here based on said
repository also has swapped IDs. This fixes the ids to reflect the
correct locations.
Signed-off-by: Evan Lojewski <github@meklort.com>
Before:
Information for the update status entry 0:
Information Version: 7
Firmware GUID: {ddc0ee61-e7f0-4e7d-acc5-c070a398838e}
Capsule Flags: 0x00000000x
Hardware Instance: 0
Update Status: attempted
Capsule File Path: /EFI/fedora/fw/fwupd-ddc0ee61-e7f0-4e7d-acc5-c070a398838e.cap
failed: Error opening file /sys/firmware/efi/efivars/fwupd-671d19d0-43c-
4852-98d9-1ce16f9967e4-0-0abba7dc-e516-4167-bbf5-4d9d1c739416: No such file
or directory
failed: Error opening file /sys/firmware/efi/efivars/fwupd-a9971959-9246-
4a5b-b2f2-ba6fdcb19349-0-0abba7dc-e516-4167-bbf5-4d9d1c739416: No such file
or directory
After:
Information for the update status entry 0:
Information Version: 7
Firmware GUID: {ddc0ee61-e7f0-4e7d-acc5-c070a398838e}
Capsule Flags: 0x00000000x
Hardware Instance: 0
Update Status: attempted
Capsule File Path: /EFI/fedora/fw/fwupd-ddc0ee61-e7f0-4e7d-acc5-c070a398838e.cap
Information for the update status entry 1:
Firmware GUID: {671d19d0-d43c-4852-98d9-1ce16f9967e4}
Update Status: No update info found
Information for the update status entry 2:
Firmware GUID: {a9971959-9246-4a5b-b2f2-ba6fdcb19349}
Update Status: No update info found
Fixes https://github.com/fwupd/fwupd/issues/2530
Although hotpluggging PCIe cards isn't common, it's how I'm testing this in the
device test enclosure. Add a tiny delay to wait for the kernel to populate the
net class on hotplug.
Unconditionally delete FWUPDATE_VERBOSE and FWUPDATE_DEBUG_LOG when deploying
the update using fwupdtool or fwupd and leave it to fwupdate.
If you want to debug the efi binary you then have to use fwupdate and squirt
the .cap file rather than using all the other layers.
Upon restarting the upstream VLI USB hub the cxaudio device re-enumerated okay
but would not service HID requests for 100ms, returning 'endpoint stalled'.
To make this more reliable retry the SetRequest up to 10 times after a short
delay to make enumeration reliable.
The GLib g_byte_array_set_size() function does not zero the contents if the
array size is larger, which leads to unpredictable output when using valgrind.
To do this without rebooting the device implement VliUsbhubPdDevice->reload()
like the other FuVliUsbhubDevice child devices.
Fixes half of https://github.com/fwupd/fwupd/issues/2376
For instance, we can tell the user that UEFI UpdateCapsule is disabled in the
system firmware, or that efivarfs is not mounted. This is much better than
creating "dummy" devices which are really just hacks around the problem because
no better API existed. THe dummy devices cause as many problems as they solve.
Plugins have to set FWUPD_PLUGIN_FLAG_USER_WARNING if a warning should be shown
to the user, and only one warning will be shown of each failure type.
It is expected that GUI clients like gnome-software and gnome-firmware would use
this API to notify the user the localized message for why firmware updates are
not being shown.
Fixes https://github.com/fwupd/fwupd/issues/2456
Add quirks for PID 5048 and 5049, which are cros-ec devices
that update via USB endpoint. Notable about these two are
they support rw_sig, for RW region signature verification.
Our update flow prior to this was:
1. On detach, reboot to RO.
2. From Old RO, Update RW.
3. Reboot to New RW
4. From New RW, Update RO.
5. Reboot to finish
This flow has a flaw, which is that the Old RO may be buggy (especially
during development, before it's actually locked down and actually Read-Only).
This change will do the following instead:
1. On detach, do not reboot to RO if RO is writeable, and we are in RW
2. From old RW, Update RO
3. Reboot to new RO
4. From New RO, update RW
5. Reboot to finish.
This has a speed advantage as well, as we save one reboot cycle (2 vs 3).
As a result of the previous change, where we send an immediate reset to
devices in order to transition back to rw, we will end up reenumerating
as RO briefly, although the destination we wanted to get to was RW.
In that case, basically skip the write firmware step, set another write
required and allow the RO firmware to complete rw_sig and jump to the rw
partition.
Previously, we sent a 'stay-in-ro' subcommand when we are in
fu_cros_ec_usb_device_reset_to_ro, which is called from the detach phase,
in other words, we are currently sitting in RW.
This is incorrect, since stay-in-ro only interrupts an RO's
rw_sig process if it is in progress.
Instead, 'stay-in-ro' must be issued when the device reenumerates in RO,
immediately before the writing sequence starts. On devices that implement
rw_sig, they will briefly enumerate as RO before self-issuing a jump to RW
on the signature check being valid. In order to stay in RO to perform a RW
partition update, we must interrupt it as soon as we see the RO version
enumerate.
- Use pci function 0 instead of 1 when flashing firmware.
In certain situations, the BCM5719 NVM controller can lockup if
a function other than 0 is used to read from NVM word-by-word like
the kernel driver does.
- Fix APE_MODE offset in BAR[2] to enable proper resetting of the APE.
- Remove unnededed NVRam lock when resetting the APE.
Signed-off-by: Evan Lojewski <github@meklort.com>
