Implement process_vm_{readv,writev}

[doanbaotrung]

Add same-process support for SYS_process_vm_readv and SYS_process_vm_writev. Import and validate guest iovec arrays, copy through guest memory in bounded chunks, and return short counts when a later buffer faults after partial progress.

Wire the syscall numbers through abi.h, dispatch.tbl, and syscall.c. Add test-process-vm coverage for scatter/gather reads and writes, short-copy fault behavior, bad flags, foreign PIDs, bad iovec pointers, oversized iovcnt, and zero-count calls. Include the test in the matrix.

Fix #197

---

Summary by cubic

Adds same-process support for SYS_process_vm_readv and SYS_process_vm_writev. Validates guest iovecs, copies in bounded chunks, and returns short counts on local/remote faults. Fixes #197.

• New Features
  • Wired syscall numbers and dispatch via abi.h, dispatch.tbl, and syscall.c.
  • Implemented iovec import/validation with length/iovcnt limits and scatter/gather copying in io.c; errors: -EINVAL for bad flags/oversized iovcnt, -ESRCH for foreign/invalid PIDs, -EFAULT for bad pointers; zero-count returns 0.
  • Added test-process-vm to the matrix covering scatter/gather reads/writes, short-copy on local/remote faults, and bad-input cases.

^{Written for commit f68a495. Summary will update on new commits.}

[cubic-dev-ai]

1 issue found across 7 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/syscall/io.c">

<violation number="1" location="src/syscall/io.c:1844">
P2: The PID gate is stricter than the rest of the runtime’s process identity handling and rejects same-process aliases outright. That can turn a same-process memory copy into `ESRCH` even though the new feature is intended to work within the current guest process.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

P2: The PID gate is stricter than the rest of the runtime’s process identity handling and rejects same-process aliases outright. That can turn a same-process memory copy into ESRCH even though the new feature is intended to work within the current guest process.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At src/syscall/io.c, line 1844:

<comment>The PID gate is stricter than the rest of the runtime’s process identity handling and rejects same-process aliases outright. That can turn a same-process memory copy into `ESRCH` even though the new feature is intended to work within the current guest process.</comment>

<file context>
@@ -1730,6 +1730,165 @@ int64_t sys_pwritev2(guest_t *g,
+{
+    if (flags != 0)
+        return -LINUX_EINVAL;
+    if (pid <= 0 || pid != proc_get_pid())
+        return -LINUX_ESRCH;
+    if (local_iovcnt == 0 || remote_iovcnt == 0)
</file context>

[doanbaotrung]

I have updated the PID validation logic inside sys_process_vm in [io.c]

1. PID Narrowing: I narrowed pid to a 32-bit signed integer (int32_t target_pid = (int32_t) pid) before validating it. This correctly addresses the low-32-bit pid_t and ignores any non-zero upper bits of the register.
2. Support for Thread IDs (Same-process Aliases): I changed the process identity check to accept the PID if it matches either proc_get_pid() (the process TGID) or any of the active thread IDs (TIDs) in the process (via thread_find()). This aligns with the rest of the runtime's permissive process-identity validation and prevents same-process memory copies from incorrectly returning ESRCH.

[jserv]

Refined based on the review from @cubic-dev-ai . If you disagree with any of its feedback, discuss it with the reviewer to reach a consensus and consolidate the shared knowledge base.

[doanbaotrung]

Fixed.
I have updated the PID validation logic inside sys_process_vm in [io.c]

1. PID Narrowing: I narrowed pid to a 32-bit signed integer (int32_t target_pid = (int32_t) pid) before validating it. This correctly addresses the low-32-bit pid_t and ignores any non-zero upper bits of the register.
2. Support for Thread IDs (Same-process Aliases): I changed the process identity check to accept the PID if it matches either proc_get_pid() (the process TGID) or any of the active thread IDs (TIDs) in the process (via thread_find()). This aligns with the rest of the runtime's permissive process-identity validation and prevents same-process memory copies from incorrectly returning ESRCH.

[jserv]

Thank @doanbaotrung for contributing!