PGL services

2024-06-11T13:21:01Z

Torphedo: Add to Services category

PGL (Program Launcher) is a sysmodule designed for encapsulating all program launching / management logic.

The core logic of this sysmodule was previously encapsulated by the ns:dev interface (and the nn.nssrv.ProcessLaunchControlTask thread in particular).

= pgl =
{| class="wikitable" border="1"
|-
! Cmd || Name
|-
| 0 || [[#LaunchProgram]]
|-
| 1 || [[#TerminateProcess]]
|-
| 2 || [[#LaunchProgramFromHost]]
|-
| 4 || [[#GetHostContentMetaInfo]]
|-
| 5 || [[#GetApplicationProcessId]]
|-
| 6 || [[#BoostSystemMemoryResourceLimit]]
|-
| 7 || [[#IsProcessTracked]]
|-
| 8 || [[#EnableApplicationCrashReport]]
|-
| 9 || [[#IsApplicationCrashReportEnabled]]
|-
| 10 || [[#EnableApplicationAllThreadDumpOnCrash]]
|-
| 12 || [[#TriggerSnapShotDumper]]
|-
| 20 || [[#GetShellEventObserver]]
|-
| 21 || [11.0.0-11.0.1]
|}

== LaunchProgram ==
Takes a [[NCM_services#ProgramLocation|ProgramLocation]], a u32 [[Process_Manager_services#LaunchFlags]] pm launch_flags, and a u8 [[#LaunchFlags|LaunchFlags]] pgl launch_flags.

This calls pm:shell->LaunchProgram with the program location and the input launch flags. If jit_debug!enable_jit_debug is true, LaunchFlags_SignalOnCrash is ORRd into the flags sent to PM.

This then creates a new process tracking structure for the new process id and with the pgl launch flags. If an empty process struct is not allocated from the 0x20 slab heap, pgl aborts.

If settings::GetDebugModeFlag() returns 1, then LaunchFlags_EnableCrashReportScreenShotForDevelop is used to enable crash report screenshot, otherwise LaunchFlags_EnableCrashReportScreenShotForProduction is used.

It returns a u64 process id.

== TerminateProcess ==
This just calls pm:shell->TerminateProcess with the input process id.

== GetShellEventObserver ==
No input. Returns an [[#IEventObserver]].

== LaunchProgramFromHost ==
This takes an input u32 [[Process_Manager_services#LaunchFlags]] pm launch_flags and a MapAlias input buffer "content path".

The content path must be to an ".nsp" file or a ".nspd" folder. "*.nspd/program0.ncd/" is also allowed.

This calls fs::MountApplicationPackage on the input content path, and reads content meta from "meta0.ncd" or ".cnmt.nca" extension folder/file depending on whether it is a .nspd or .nsp content path.

A location resolver for the Host storage is opened, and EraseProgramRedirection is called for the parsed program id.

RedierectProgramPath is then called for the program ID to set it to the input content path.

This then calls the implementation function for [[#LaunchProgram]] with the input pm launch flags and pgl launch flags = 0.

== GetHostContentMetaInfo ==
This takes an input MapAlias buffer "content path", and does the same parsing as [[#LaunchProgramFromHost]].

Instead of launching the program, it instead returns an 0x10 output [[#ContentMetaInfo]] from the parsed info.

== GetApplicationProcessId ==
This calls pm:shell->GetApplicationProcessIdForShell, and returns the output process id.

Error 0x6E4 is returned when there is no application process.

== BoostSystemMemoryResourceLimit ==
This takes in a u64 mem_size, and uses it as input to pm:shell->BoostSystemMemoryResourceLimit

== EnableApplicationCrashReport ==
Takes an input boolean "enable".

This calls pm:shell->GetApplicationProcessId. If the returned process ID is tracked, this sets or clears bit 1 (value 0x2) in the tracked process's flags.

When a process crash event occurs and pgl launches crash report, the second argument is "%d" formatted with the bitwise AND of this bit and bit 0 (value 0x1).

LaunchFlags_EnableDetailedCrashReport causes the process to have both of these flags set, otherwise they are both zero.

Thus this command is only useful for disabling/toggling detailed crash reports for applications launched with LaunchFlags_EnableDetailedCrashReport, as processes without the launch flag will always send "0" for the creport argument.

== IsApplicationCrashReportEnabled ==
This returns the value of the bit set or cleared by [[#EnableApplicationCrashReport]]

== EnableApplicationAllThreadDumpOnCrash ==
Takes an input boolean "enable".

This calls pm:shell->GetApplicationProcessId. If the returned process ID is tracked, this sets or clears bit 3 (value 0x8) in the tracked process's flags.

Using this command unconditionally sets bit 2 (value 0x4) in the tracked process's flags.

When the SnapShotDumper is launched, argument "-log %d" is set to the value of bit 3 if bit 2 is 1. Otherwise, argument "-log %d" is set to the value of the system setting "snap_shot_dump!output_all_log".

Thus, this overrides the log option for the current application process from snap_shot_dump!output_all_log with the input value.

Previously this functionality was controlled by ns:am2->EnableApplicationAllThreadDumpOnCrash, which is now a wrapper for this function.

== TriggerSnapShotDumper ==
This takes an input 32 "SnapShotDumpType" and an input MapAliasBuffer "argument".

This manually invokes the same function to launch the SnapShotDumper program (0100000000002071) as is called automatically when a process crashes and jit_debug!enable_jit_debug is 1.

If the argument is NULL, then the snap shot dumper is launched with argument Format("D %010llu -log %d -dump %d", process_id, log_option, snap_shot_dump_type - 1).
If the argument is not NULL, the snap shot dumper is launched with argument Format("D %010llu \"%s\" -log %d -dump %d", process_id, argument, log_option, snap_shot_dump_type - 1).

When invoked automatically, the SnapShotDumpType is SnapShotDumpType_Full if both "snap_shot_dump!full_dump" and "snap_shot_dump!auto_dump" are true, SnapShotDumpType_Auto if only "snap_shot_dump!auto_dump" is 1, and SnapShotDumpType_None otherwise.

== IEventObserver ==
{| class="wikitable" border="1"
|-
! Cmd || Name
|-
| 0 || [[#GetProcessEventHandle]]
|-
| 1 || [[#GetProcessEventInfo]]
|}

== GetProcessEventHandle ==
This takes no input and returns an output copy event handle.

== GetProcessEventInfo ==
This takes no input and attempts to receive a [[Process_Manager_services#ProcessEventInfo]] from the observer's message queue.

Every time a process event occurs, pgl receives the event from PM (it does pm:shell->GetProcessEventHandle() during startup and has a thread looping pm:shell->GetProcessEventInfo() to track process events.

PGL sends to all observers' message queues a [[Process_Manager_services#ProcessEventInfo]] for each event it receives, thus allowing for multiple subscribers to process events.

= LaunchFlags =
{| class=wikitable
! Bit || Mask || Name
|-
| 0 || 1 || EnableDetailedCrashReport
|-
| 1 || 2 || EnableCrashReportScreenShotForProduction
|-
| 2 || 4 || EnableCrashReportScreenShotForDevelop
|-
|}

= SnapShotDumpType =
{| class=wikitable
! Value || Name
|-
| 0 || None
|-
| 1 || Auto
|-
| 2 || Full
|-
|}

= ContentMetaInfo =
{| class="wikitable" border="1"
|-
! Offset
! Size
! Description
|-
| 0x0
| 0x8
| [[NCM_services#ProgramId|ProgramId]]
|-
| 0x8
| 0x4
| Version
|-
| 0xC
| 0x1
| ContentType
|-
| 0xD
| 0x1
| ID Offset
|-
| 0xE
| 0x2
| Padding
|}

[[Category:Services]]

Tegra X1

2024-03-05T23:01:19Z

Torphedo: The google drive link denies access. This was surprisingly annoying to track down

NVIDIA Tegra X1 SoC Specifications

CPU
* Quad 64-bit A57 cores + Quad 64-bit A53 cores

Cache
* Cortex A57 cluster: 2 MB Shared L2 Cache, 48KB /32KB (I/D) L1 Cache per core
* Cortex A53 cluster: 512KB shared L2 Cache, 32KB/32KB (I/D) L1 Cache per core

Memory Frequency
* LPDDR3, LPDDR4-1600, 64-bit (25.6 GB/s)

Memory Size
* Up to 4 GB

GPU
* Cores 256-core Maxwell GPU with support for FP16
* API Support OpenGL ES 3.1, OpenGL4.5, DirectX 12.0, AEP, CUDA 6.0

Video
* Decode VP9, H.265, H.264 4K 60 fps; H.265 4K 60fps 10-bit color; VP8 1080p 60fps;
* Encode H.264, H.265 4K 30 fps; VP8 1080p 60 fps;

Imaging
* Image Processing Dual ISP, 1.3 GigaPixels/s, 4096 focus points, 100 MP Sensor support, up to 6 camera inputs
* JPEG Decode/Encode 600 MPixels/s

Display
* Display Controllers 2 Simultaneous
* HDMI HDMI 2.0, HDCP 2.2, 4K 60 fps
* Local Display 4K 60 fps VESA DSC compression

Storage
* Storage interface e-MMC 5.1 (HS533), CMD Queuing

Tegra X1 TRM datasheet
* https://developer.nvidia.com/embedded/downloads#?search=X1 (2977 pages, requires free Nvidia developer account)
* https://drive.google.com/file/d/0B5cebp2-yLm5YURqeFROb2hwOUk/view

Nintendo Switch Brew - User contributions [en]

PGL services

Tegra X1