ladybird/Kernel/Bus/PCI/Access.cpp

274 lines
11 KiB
C++
Raw Normal View History

/*
2020-03-06 11:59:29 -03:00
* Copyright (c) 2020, Liav A. <liavalb@hotmail.co.il>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
#include <AK/ByteReader.h>
#include <AK/Error.h>
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
#include <AK/HashTable.h>
2022-10-03 21:05:54 -03:00
#if ARCH(X86_64)
# include <Kernel/Arch/x86_64/PCI/Controller/HostBridge.h>
#endif
#include <Kernel/Bus/PCI/Access.h>
#include <Kernel/Bus/PCI/Controller/MemoryBackedHostBridge.h>
#include <Kernel/Bus/PCI/Initializer.h>
Meta: Split debug defines into multiple headers. The following script was used to make these changes: #!/bin/bash set -e tmp=$(mktemp -d) echo "tmp=$tmp" find Kernel \( -name '*.cpp' -o -name '*.h' \) | sort > $tmp/Kernel.files find . \( -path ./Toolchain -prune -o -path ./Build -prune -o -path ./Kernel -prune \) -o \( -name '*.cpp' -o -name '*.h' \) -print | sort > $tmp/EverythingExceptKernel.files cat $tmp/Kernel.files | xargs grep -Eho '[A-Z0-9_]+_DEBUG' | sort | uniq > $tmp/Kernel.macros cat $tmp/EverythingExceptKernel.files | xargs grep -Eho '[A-Z0-9_]+_DEBUG' | sort | uniq > $tmp/EverythingExceptKernel.macros comm -23 $tmp/Kernel.macros $tmp/EverythingExceptKernel.macros > $tmp/Kernel.unique comm -1 $tmp/Kernel.macros $tmp/EverythingExceptKernel.macros > $tmp/EverythingExceptKernel.unique cat $tmp/Kernel.unique | awk '{ print "#cmakedefine01 "$1 }' > $tmp/Kernel.header cat $tmp/EverythingExceptKernel.unique | awk '{ print "#cmakedefine01 "$1 }' > $tmp/EverythingExceptKernel.header for macro in $(cat $tmp/Kernel.unique) do cat $tmp/Kernel.files | xargs grep -l $macro >> $tmp/Kernel.new-includes ||: done cat $tmp/Kernel.new-includes | sort > $tmp/Kernel.new-includes.sorted for macro in $(cat $tmp/EverythingExceptKernel.unique) do cat $tmp/Kernel.files | xargs grep -l $macro >> $tmp/Kernel.old-includes ||: done cat $tmp/Kernel.old-includes | sort > $tmp/Kernel.old-includes.sorted comm -23 $tmp/Kernel.new-includes.sorted $tmp/Kernel.old-includes.sorted > $tmp/Kernel.includes.new comm -13 $tmp/Kernel.new-includes.sorted $tmp/Kernel.old-includes.sorted > $tmp/Kernel.includes.old comm -12 $tmp/Kernel.new-includes.sorted $tmp/Kernel.old-includes.sorted > $tmp/Kernel.includes.mixed for file in $(cat $tmp/Kernel.includes.new) do sed -i -E 's/#include <AK\/Debug\.h>/#include <Kernel\/Debug\.h>/' $file done for file in $(cat $tmp/Kernel.includes.mixed) do echo "mixed include in $file, requires manual editing." done
2021-01-25 12:07:10 -03:00
#include <Kernel/Debug.h>
#include <Kernel/Firmware/ACPI/Definitions.h>
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
#include <Kernel/Memory/MemoryManager.h>
#include <Kernel/Memory/Region.h>
#include <Kernel/Memory/TypedMapping.h>
#include <Kernel/Sections.h>
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
namespace Kernel::PCI {
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
#define PCI_MMIO_CONFIG_SPACE_SIZE 4096
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
static Access* s_access;
Access& Access::the()
{
if (s_access == nullptr) {
VERIFY_NOT_REACHED(); // We failed to initialize the PCI subsystem, so stop here!
}
return *s_access;
}
bool Access::is_initialized()
{
return (s_access != nullptr);
}
bool Access::is_hardware_disabled()
{
return g_pci_access_io_probe_failed;
}
bool Access::is_disabled()
{
return g_pci_access_is_disabled_from_commandline || g_pci_access_io_probe_failed;
}
UNMAP_AFTER_INIT bool Access::find_and_register_pci_host_bridges_from_acpi_mcfg_table(PhysicalAddress mcfg_table)
{
u32 length = 0;
u8 revision = 0;
{
auto mapped_mcfg_table_or_error = Memory::map_typed<ACPI::Structures::SDTHeader>(mcfg_table);
if (mapped_mcfg_table_or_error.is_error()) {
dbgln("Failed to map MCFG table");
return false;
}
auto mapped_mcfg_table = mapped_mcfg_table_or_error.release_value();
length = mapped_mcfg_table->length;
revision = mapped_mcfg_table->revision;
}
if (length == sizeof(ACPI::Structures::SDTHeader))
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
return false;
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
dbgln("PCI: MCFG, length: {}, revision: {}", length, revision);
if (Checked<size_t>::addition_would_overflow(length, PAGE_SIZE)) {
dbgln("Overflow when adding extra page to allocation of length {}", length);
return false;
}
length += PAGE_SIZE;
auto region_size_or_error = Memory::page_round_up(length);
if (region_size_or_error.is_error()) {
dbgln("Failed to round up length of {} to pages", length);
return false;
}
auto mcfg_region_or_error = MM.allocate_kernel_region(mcfg_table.page_base(), region_size_or_error.value(), "PCI Parsing MCFG"sv, Memory::Region::Access::ReadWrite);
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
if (mcfg_region_or_error.is_error())
return false;
auto& mcfg = *(ACPI::Structures::MCFG*)mcfg_region_or_error.value()->vaddr().offset(mcfg_table.offset_in_page()).as_ptr();
dbgln_if(PCI_DEBUG, "PCI: Checking MCFG @ {}, {}", VirtualAddress(&mcfg), mcfg_table);
for (u32 index = 0; index < ((mcfg.header.length - sizeof(ACPI::Structures::MCFG)) / sizeof(ACPI::Structures::PCI_MMIO_Descriptor)); index++) {
u8 start_bus = mcfg.descriptors[index].start_pci_bus;
u8 end_bus = mcfg.descriptors[index].end_pci_bus;
u64 start_addr = mcfg.descriptors[index].base_addr;
Domain pci_domain { index, start_bus, end_bus };
dmesgln("PCI: New PCI domain @ {}, PCI buses ({}-{})", PhysicalAddress { start_addr }, start_bus, end_bus);
auto host_bridge = MemoryBackedHostBridge::must_create(pci_domain, PhysicalAddress { start_addr });
add_host_controller(move(host_bridge));
}
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
return true;
}
UNMAP_AFTER_INIT bool Access::initialize_for_multiple_pci_domains(PhysicalAddress mcfg_table)
{
VERIFY(!Access::is_initialized());
auto* access = new Access();
if (!access->find_and_register_pci_host_bridges_from_acpi_mcfg_table(mcfg_table))
return false;
access->rescan_hardware();
dbgln_if(PCI_DEBUG, "PCI: access for multiple PCI domain initialised.");
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
return true;
}
2022-10-03 21:05:54 -03:00
#if ARCH(X86_64)
UNMAP_AFTER_INIT bool Access::initialize_for_one_pci_domain()
{
VERIFY(!Access::is_initialized());
auto* access = new Access();
auto host_bridge = HostBridge::must_create_with_io_access();
access->add_host_controller(move(host_bridge));
access->rescan_hardware();
dbgln_if(PCI_DEBUG, "PCI: access for one PCI domain initialised.");
return true;
}
#endif
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
ErrorOr<void> Access::add_host_controller_and_scan_for_devices(NonnullOwnPtr<HostController> controller)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
SpinlockLocker locker(m_access_lock);
SpinlockLocker scan_locker(m_scan_lock);
auto domain_number = controller->domain_number();
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(!m_host_controllers.contains(domain_number));
// Note: We need to register the new controller as soon as possible, and
// definitely before enumerating devices behind that.
m_host_controllers.set(domain_number, move(controller));
ErrorOr<void> error_or_void {};
m_host_controllers.get(domain_number).value()->enumerate_attached_devices([&](EnumerableDeviceIdentifier const& device_identifier) -> IterationDecision {
auto device_identifier_or_error = DeviceIdentifier::from_enumerable_identifier(device_identifier);
if (device_identifier_or_error.is_error()) {
error_or_void = device_identifier_or_error.release_error();
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
return IterationDecision::Break;
}
m_device_identifiers.append(device_identifier_or_error.release_value());
return IterationDecision::Continue;
});
return {};
}
UNMAP_AFTER_INIT void Access::add_host_controller(NonnullOwnPtr<HostController> controller)
{
auto domain_number = controller->domain_number();
m_host_controllers.set(domain_number, move(controller));
}
UNMAP_AFTER_INIT Access::Access()
{
s_access = this;
}
UNMAP_AFTER_INIT void Access::rescan_hardware()
{
SpinlockLocker locker(m_access_lock);
SpinlockLocker scan_locker(m_scan_lock);
VERIFY(m_device_identifiers.is_empty());
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
ErrorOr<void> error_or_void {};
for (auto it = m_host_controllers.begin(); it != m_host_controllers.end(); ++it) {
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
(*it).value->enumerate_attached_devices([this, &error_or_void](EnumerableDeviceIdentifier device_identifier) -> IterationDecision {
auto device_identifier_or_error = DeviceIdentifier::from_enumerable_identifier(device_identifier);
if (device_identifier_or_error.is_error()) {
error_or_void = device_identifier_or_error.release_error();
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
return IterationDecision::Break;
}
m_device_identifiers.append(device_identifier_or_error.release_value());
return IterationDecision::Continue;
});
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
if (error_or_void.is_error()) {
dmesgln("Failed during PCI Access::rescan_hardware due to {}", error_or_void.error());
VERIFY_NOT_REACHED();
}
}
ErrorOr<void> Access::fast_enumerate(Function<void(DeviceIdentifier const&)>& callback) const
{
// Note: We hold the m_access_lock for a brief moment just to ensure we get
// a complete Vector in case someone wants to mutate it.
Vector<NonnullRefPtr<DeviceIdentifier>> device_identifiers;
{
SpinlockLocker locker(m_access_lock);
VERIFY(!m_device_identifiers.is_empty());
TRY(device_identifiers.try_extend(m_device_identifiers));
}
for (auto const& device_identifier : device_identifiers) {
callback(device_identifier);
}
return {};
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
DeviceIdentifier const& Access::get_device_identifier(Address address) const
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
for (auto& device_identifier : m_device_identifiers) {
auto device_address = device_identifier->address();
if (device_address.domain() == address.domain()
&& device_address.bus() == address.bus()
&& device_address.device() == address.device()
&& device_address.function() == address.function()) {
return device_identifier;
}
}
VERIFY_NOT_REACHED();
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
void Access::write8_field(DeviceIdentifier const& identifier, u32 field, u8 value)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
controller.write8_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field, value);
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
void Access::write16_field(DeviceIdentifier const& identifier, u32 field, u16 value)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
controller.write16_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field, value);
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
void Access::write32_field(DeviceIdentifier const& identifier, u32 field, u32 value)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
controller.write32_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field, value);
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
u8 Access::read8_field(DeviceIdentifier const& identifier, RegisterOffset field)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
return read8_field(identifier, to_underlying(field));
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
u16 Access::read16_field(DeviceIdentifier const& identifier, RegisterOffset field)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
return read16_field(identifier, to_underlying(field));
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
u8 Access::read8_field(DeviceIdentifier const& identifier, u32 field)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
return controller.read8_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field);
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
u16 Access::read16_field(DeviceIdentifier const& identifier, u32 field)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
return controller.read16_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field);
}
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
u32 Access::read32_field(DeviceIdentifier const& identifier, u32 field)
{
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(identifier.operation_lock().is_locked());
SpinlockLocker locker(m_access_lock);
Kernel/PCI: Hold a reference to DeviceIdentifier in the Device class There are now 2 separate classes for almost the same object type: - EnumerableDeviceIdentifier, which is used in the enumeration code for all PCI host controller classes. This is allowed to be moved and copied, as it doesn't support ref-counting. - DeviceIdentifier, which inherits from EnumerableDeviceIdentifier. This class uses ref-counting, and is not allowed to be copied. It has a spinlock member in its structure to allow safely executing complicated IO sequences on a PCI device and its space configuration. There's a static method that allows a quick conversion from EnumerableDeviceIdentifier to DeviceIdentifier while creating a NonnullRefPtr out of it. The reason for doing this is for the sake of integrity and reliablity of the system in 2 places: - Ensure that "complicated" tasks that rely on manipulating PCI device registers are done in a safe manner. For example, determining a PCI BAR space size requires multiple read and writes to the same register, and if another CPU tries to do something else with our selected register, then the result will be a catastrophe. - Allow the PCI API to have a united form around a shared object which actually holds much more data than the PCI::Address structure. This is fundamental if we want to do certain types of optimizations, and be able to support more features of the PCI bus in the foreseeable future. This patch already has several implications: - All PCI::Device(s) hold a reference to a DeviceIdentifier structure being given originally from the PCI::Access singleton. This means that all instances of DeviceIdentifier structures are located in one place, and all references are pointing to that location. This ensures that locking the operation spinlock will take effect in all the appropriate places. - We no longer support adding PCI host controllers and then immediately allow for enumerating it with a lambda function. It was found that this method is extremely broken and too much complicated to work reliably with the new paradigm being introduced in this patch. This means that for Volume Management Devices (Intel VMD devices), we simply first enumerate the PCI bus for such devices in the storage code, and if we find a device, we attach it in the PCI::Access method which will scan for devices behind that bridge and will add new DeviceIdentifier(s) objects to its internal Vector. Afterwards, we just continue as usual with scanning for actual storage controllers, so we will find a corresponding NVMe controllers if there were any behind that VMD bridge.
2022-02-10 13:33:13 -03:00
VERIFY(m_host_controllers.contains(identifier.address().domain()));
auto& controller = *m_host_controllers.get(identifier.address().domain()).value();
return controller.read32_field(identifier.address().bus(), identifier.address().device(), identifier.address().function(), field);
}
Kernel/PCI: Simplify the entire subsystem A couple of things were changed: 1. Semantic changes - PCI segments are now called PCI domains, to better match what they are really. It's also the name that Linux gave, and it seems that Wikipedia also uses this name. We also remove PCI::ChangeableAddress, because it was used in the past but now it's no longer being used. 2. There are no WindowedMMIOAccess or MMIOAccess classes anymore, as they made a bunch of unnecessary complexity. Instead, Windowed access is removed entirely (this was tested, but never was benchmarked), so we are left with IO access and memory access options. The memory access option is essentially mapping the PCI bus (from the chosen PCI domain), to virtual memory as-is. This means that unless needed, at any time, there is only one PCI bus being mapped, and this is changed if access to another PCI bus in the same PCI domain is needed. For now, we don't support mapping of different PCI buses from different PCI domains at the same time, because basically it's still a non-issue for most machines out there. 2. OOM-safety is increased, especially when constructing the Access object. It means that we pre-allocating any needed resources, and we try to find PCI domains (if requested to initialize memory access) after we attempt to construct the Access object, so it's possible to fail at this point "gracefully". 3. All PCI API functions are now separated into a different header file, which means only "clients" of the PCI subsystem API will need to include that header file. 4. Functional changes - we only allow now to enumerate the bus after a hardware scan. This means that the old method "enumerate_hardware" is removed, so, when initializing an Access object, the initializing function must call rescan on it to force it to find devices. This makes it possible to fail rescan, and also to defer it after construction from both OOM-safety terms and hotplug capabilities.
2021-09-07 06:08:38 -03:00
}