Commit Graph

154 Commits

Author SHA1 Message Date
Jacob Lorentzon e7a6bb6c81 Only do one memory:physical map per MCFG entry. 2024-02-20 22:27:14 +00:00
Jeremy Soller de2fae8183 Do not compile MSI-X function for x86 32-bit 2024-02-16 20:52:37 -07:00
bjorn3 67d3015e2e Add helper for allocating a single MSI/MSI-X interrupt vector
It doesn't actually configure the device to emit it though, but does
make this easier to do.
2024-01-25 14:43:44 +01:00
bjorn3 ecdcefba69 Clarify message_address argument names 2024-01-24 14:28:40 +01:00
bjorn3 f43aa8574c Simplify MSI-X table pointer computation 2024-01-24 14:28:40 +01:00
bjorn3 7dbf22331b Share msix config validation code 2024-01-24 14:28:40 +01:00
bjorn3 354c351b19 Unify device info printing between drivers 2024-01-23 13:01:40 +01:00
bjorn3 77b97bb2ed Introduce PciBar::physmap_mem helper 2024-01-23 12:02:40 +01:00
bjorn3 4302a35a7e Remove a couple of unused functions from MsixCapability
The set and write methods were useless as no rpc call is exposed by
pcid to actually write the new values. As for the rest, there were
no uses and they can be emulated using the other methods.
2024-01-22 20:44:17 +01:00
bjorn3 ad23ad5d93 Introduce irq_handle helper for legacy interrupts 2024-01-22 19:35:18 +01:00
bjorn3 bd6716efe4 Make legacy_interrupt_line an Option
Drivers don't need to know the exact legacy_interrupt_pin in use.
Just if legacy interrupts are enabled and if so which
legacy_interrupt_line is used.
2024-01-22 19:35:06 +01:00
bjorn3 e972047567 Use EndpointHeader methods for the command and status registers 2024-01-22 18:12:37 +01:00
bjorn3 623d12f276 Only try to load drivers for endpoint functions
For bridge functions we don't have any existing drivers and we
currently panic if we would try to load a driver for them.
2024-01-22 18:03:46 +01:00
bjorn3 545596f296 Don't try to fetch BARs unless there is a matching driver
Fetching BARs will temporarily invalidate them, which could cause
crashes if a driver happens to access it at the same time. As we
don't yet use a single pcid instance, we don't know which devices
have a driver attached or not. As such approximate this as the set
of devices for which the current pcid instance would load a driver.
2024-01-22 17:54:38 +01:00
bjorn3 18529f6cee Remove PciClass in favor of pci_types::DeviceType 2024-01-22 16:57:53 +01:00
bjorn3 fb0dcb384b Use pci_types in a couple more places in PciHeader::from_reader 2024-01-22 16:48:52 +01:00
bjorn3 d2431d4140 Make all parsing in PciHeader::from_reader use pci_types 2024-01-22 16:32:18 +01:00
bjorn3 a5d0c7c354 Use pci_types for reading BARs
This simplifies a lot of code and adds support for 64bit BARs.
2024-01-22 15:54:18 +01:00
bjorn3 d1b6009e3d Start converting PciHeader::from_reader to pci_types 2024-01-22 15:25:08 +01:00
bjorn3 0b611dca04 Start using the pci_types crate 2024-01-22 11:25:43 +01:00
bjorn3 d56881de88 Use CfgAccess instead of ConfigReader in PciHeader::from_reader
This will enable getting BAR sizes directly inside this function in the future.
2024-01-22 11:00:17 +01:00
bjorn3 cc015eab13 Move PciHeader out of the pcid_interface crate 2024-01-21 21:25:50 +01:00
bjorn3 5192816de7 Introduce expect_port and expect_mem helpers 2024-01-21 21:12:24 +01:00
bjorn3 be0e6b9dd7 Pass self instead of &self to PciFeature methods
PciFeature is Copy, so passing it behind a reference is not necessary.
2024-01-21 20:46:00 +01:00
bjorn3 d0e90f5ded Remove fetch_header from pcid_interface
Most things are already stored in SubdriverArguments and the couple of
things that aren't may change over time and thus should be read again
each time they are accessed, while fetch_header would receive a fixed
copy from pcid.
2024-01-21 19:35:39 +01:00
bjorn3 e5772c132b Remove a lot of fields from PciHeader which are unlikely to be used
Some of these are removed with PCIe while many others only need to be
used by the firmware to initialize the PCI bridges. If we ever need
them anyway, we can always add them back, but for now it improves
readability.
2024-01-21 19:12:35 +01:00
bjorn3 1890293e86 Introduce a FullDeviceId type and pass it in pcid_interface 2024-01-21 15:18:56 +01:00
bjorn3 92914e808c Reduce code indentation one level 2024-01-21 14:03:33 +01:00
bjorn3 b9c4c61dcc Move driver matching to config.rs 2024-01-21 13:02:47 +01:00
bjorn3 92428c535b Only enable redox scheme logger when compiling for redox
This allows running pcid tests on the host.
2024-01-21 11:35:06 +01:00
bjorn3 dc13752d74 Force usage of pcid_interface 2024-01-20 19:42:06 +01:00
bjorn3 382047b14e Rename the pcie module to cfg_access 2024-01-20 16:04:18 +01:00
bjorn3 d832717bf9 Remove a bit of dead code 2024-01-20 14:42:29 +01:00
bjorn3 20eb92ae02 Assert PCI config space accesses are aligned 2024-01-20 14:36:37 +01:00
bjorn3 cb7dc2abd2 Only set iopl for helper threads when actually necessary 2024-01-20 14:17:51 +01:00
bjorn3 3cbfbf6442 Move the IO port based fallback for PCI out of the pci module
The pci module is included in the pcid_interface library, which never
needs it.
2024-01-20 14:07:31 +01:00
bjorn3 4f20b90fc3 Remove PciBus and PciDev types
They weren't used in a load bearing way. If we want to keep persistent
state about buses and devices in the future they would probably get a
different shape.
2024-01-20 13:38:45 +01:00
bjorn3 cf7fbacdaf Stop leaking Mcfg again
We don't need it after Pcie::new anymore
2024-01-19 19:53:39 +01:00
bjorn3 529b4935ee Physmap all buses upfront
This avoids a costly Mutex lock and BTreeMap lookup for each config
space access.
2024-01-19 19:50:11 +01:00
bjorn3 23d963361a Fix a couple of warnings 2024-01-19 19:14:22 +01:00
bjorn3 cb9edcb7d6 Reduce log verbosity of Capability::parse_vendor by combining some logs 2024-01-19 19:11:05 +01:00
bjorn3 8d87b701ba Remove Mcfgs type
This wrapper was only useful back when we accepted multiple MCFG tables.
2024-01-19 19:10:35 +01:00
bjorn3 ae7e2dac9f Only parse a single MCFG ACPI table
There should only be one and Linux only parses the first one too.
2024-01-19 18:41:28 +01:00
bjorn3 07c029ec21 Add a fixme for using ACPI 2024-01-19 18:10:57 +01:00
bjorn3 101f062b56 Remove a couple of error conditions in Mcfgs::fetch()
* Every directory entry has to have a filename, so use unwrap
* Handle non-UTF-8 ACPI table filenames by using as_encoded_bytes
2024-01-19 14:58:09 +01:00
bjorn3 240e59e2bf Handle pci buses without a device at the first slot 2024-01-19 14:45:48 +01:00
bjorn3 c611988768 Fix MCFG parsing for PCIe 2024-01-19 13:19:53 +01:00
bjorn3 b317484d67 Simplify Mcfgs::fetch 2024-01-19 13:00:23 +01:00
bjorn3 15249127a5 Fix a couple of warnings 2024-01-19 12:47:42 +01:00
bjorn3 d010d03eef Fix pcid tests 2024-01-19 12:40:55 +01:00