Development:32Bit Syscall Woes - Revision history

Sonicadvance1 at 00:33, 1 December 2025

2025-12-01T00:33:27Z

2025-08-15T05:57:24Z

2023-10-21T20:06:32Z

Robust list/futex family

2023-10-21T20:06:05Z

Robust list/futex family

2023-10-21T20:05:47Z

Robust list/futex family

2023-10-18T07:43:09Z

2023-08-21T16:56:06Z

2023-08-21T16:38:06Z

Robust list/futex family

2022-07-29T21:24:35Z

2022-07-29T21:20:59Z

← Older revision		Revision as of 00:33, 1 December 2025
Line 293:		Line 293:
	* https://github.com/FEX-Emu/FEX/pull/4106		* https://github.com/FEX-Emu/FEX/pull/4106
	* https://github.com/slp/linux/commit/897560fd00e1add63a9973c31fc437b006855680		* https://github.com/slp/linux/commit/897560fd00e1add63a9973c31fc437b006855680
		+
		+	== io_uring - '''Happens on x86-64 as well''' ==
		+	The queues can have `epoll_event` objects encoded in them. These are packed 12-byte structs on x86/x86-64, but 16-byte on other architectures. We can't reasonably re-encode these user-queues.
		+
		+	=== Possible fixes ===
		+	A flag to io_uring setup that epoll_event must be packed like x86?
		+
		+	=== Current Implementation ===
		+	Disabled

← Older revision		Revision as of 05:57, 15 August 2025
Line 285:		Line 285:
	=== Possible fixes ===		=== Possible fixes ===
	TBD		TBD
		+
		+	== Read/Write syscalls providing different data depending on process bitness ==
		+	Means gamepad input/output is completely broken for 32-bit emulated processes.
		+
		+	=== Current implementation ===
		+	Just be broken, or downstream patches to enable "compat" read/write
		+	* https://github.com/FEX-Emu/FEX/pull/4106
		+	* https://github.com/slp/linux/commit/897560fd00e1add63a9973c31fc437b006855680

@@ Line 229: / Line 229: @@
 === Possible fixes ===
 * Implement a new key based robust list syscall
 ** Instead of just having a single robust list per process (technically two for x86-64)

@@ Line 249: / Line 249: @@
    };
    void* robust_list_ctl(robust_list_op op, ...);
-  eg:
    // Allocate a robust list object
    int robust_list_ctl(OP_ALLOC); // Returning a key object on success. Default NULL list head.

@@ Line 258: / Line 258: @@
    // Register a robust futex list to an object.
    // Taking a 64-bit robust futex linked list just like set_robust_list
    int robust_list_ctl(OP_REGISTER, <int key>, TYPE_ROBUST_64BIT, <struct robust_list_head **head_ptr>);
@@ Line 263: / Line 264: @@
    // Taking a 32-bit pointer packed robust futex list.
    // Just like 32-bit compat set_robust_list
    int robust_list_ctl(OP_REGISTER, <int key>, TYPE_ROBUST_32BIT, <compat_ptr_list(struct robust_list_head **)head_ptr>);

@@ Line 252: / Line 252: @@
    // Allocate a robust list object
    int robust_list_ctl(OP_ALLOC); // Returning a key object on success. Default NULL list head.
    // Free a robust list object
    int robust_list_ctl(OP_FREE, <int key>); // Taking a key object as an argument. Error on invalid key. Removes robust list from kernel tracking.
    // Register a robust futex list to an object.
    // Taking a 64-bit robust futex linked list just like set_robust_list
    int robust_list_ctl(OP_REGISTER, <int key>, TYPE_ROBUST_64BIT, <struct robust_list_head **head_ptr>);
    // Register a compat capable robust futex list to an object
    // Taking a 32-bit pointer packed robust futex list.
    // Just like 32-bit compat set_robust_list
    int robust_list_ctl(OP_REGISTER, <int key>, TYPE_ROBUST_32BIT, <compat_ptr_list(struct robust_list_head **)head_ptr>);
    // Get the current pointer that a robust_futex_object points to. Just like get_robust_list
    // Returns either a 32-bit or 64-bit head_ptr depending on registered type.
    <struct robust_list_head **head_ptr> robust_list_ctl(OP_GET, <int key>);
 === Current implementation ===
 * 64-bit processes

@@ Line 219: / Line 219: @@
 === Current Implementation ===
-* getdents - return -ENOSYS and praying nothing uses it
+* getdents - Emulating with getdents and rewriting all elements
 * getdents64 - overwriting d_off to be a incrementing number and hoping nothing uses it
@@ Line 238: / Line 238: @@
 ** This way the kernel will know if the passed in linked list should be tracked using 32-bit pointers or 64-bit pointers.
 === Current implementation ===
 * 64-bit processes

@@ Line 241: / Line 241: @@
 * 64-bit processes
 ** Pass the robust list through so the kernel is tracking the guest application's robust list
 * 32-bit processes
 ** Lie to them and say it is getting tracked, won't actually work.

@@ Line 168: / Line 168: @@
 * MAP_GROWSUP and MAP_GROWSDOWN
-=== IOCTL number conflicts are non-trivial to resolve in userspace ====
+=== IOCTL number conflicts are non-trivial to resolve in userspace ===
 * Just take a look at ioctl-number.rst, there are 123 instances of conflict in there
 * Trying to resolve conflicts back to the correct device? How?
@@ Line 176: / Line 176: @@
 === Current implementation ===
-See mmap/mmap2 current implementation
+See mmap/mmap2 current implementation to handle memory allocation problems.
 == sendmsg, recvmsg, sendmmsg, recvmmsg - (socketcall) 102, 370, 372, 345, 337/417 ==

@@ Line 186: / Line 186: @@
 This potential footgun on the application facing side turns in to an even larger footgun on the FEX side.
-== Current Implementation ==
+=== Current Implementation ===
 Try to patch up the data as well as we can and just eat the fact that guest applications could accidentally read too much data and lose it.