syscalls: beef up write(2) and pwrite(2) tests #12426

0xzre · 2026-01-03T14:49:18Z

Add comprehensive tests for write(2), pwrite(2), and writev(2) syscalls
to improve coverage as requested in the issue.

New tests include:

Bad buffer states (EFAULT): nullptr buffer, bad iov_base address
Bad file descriptors (EBADF): closed fd, negative fd, read-only fd, O_PATH fd
Offset behavior: verify offset increments, pwrite doesn't change offset
EOF handling: write extends past EOF, pwrite creates holes
Pipes: write to pipe, pwrite returns ESPIPE, write to read end fails
Symlinks: write through symlink, O_NOFOLLOW returns ELOOP
Special files: /dev/null and /dev/zero
writev edge cases: empty iovec, zero-length entries

Also adds similar tests to pwrite64.cc for consistency.

google-cla · 2026-01-03T14:49:24Z

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

test/syscalls/linux/write.cc

test/syscalls/linux/pwrite64.cc

ayushr2 · 2026-01-06T15:30:10Z

Thanks, could you squash your commits?

gvisor/CONTRIBUTING.md

Lines 95 to 99 in 4e9165c

    
           When iterating on a pull request, please amend your commit rather than creating 
        
           new commits on the same branch. Our merge bot applies all commits from the 
        
           branch to master and does not yet support squash merging. If you'd like to 
        
           submit a pull request with multiple commits, please ensure they are independent, 
        
           logical units of work that can be applied separately.

0xzre · 2026-01-06T15:42:08Z

thanks for the guidance @ayushr2 , squashed them🙏

Fixes #2370 Add comprehensive tests for write(2), pwrite(2), and writev(2) syscalls to improve coverage as requested in the issue. New tests include: - Bad buffer states (EFAULT): nullptr buffer, bad iov_base address - Bad file descriptors (EBADF): closed fd, negative fd, read-only fd, O_PATH fd - Offset behavior: verify offset increments, pwrite doesn't change offset - EOF handling: write extends past EOF, pwrite creates holes - Pipes: write to pipe, pwrite returns ESPIPE, write to read end fails - Symlinks: write through symlink, O_NOFOLLOW returns ELOOP - Special files: /dev/null and /dev/zero - writev edge cases: empty iovec, zero-length entries Also adds similar tests to `pwrite64.cc` for consistency. FUTURE_COPYBARA_INTEGRATE_REVIEW=#12426 from 0xzre:write-syscall-tests 2b30584 PiperOrigin-RevId: 852782177

0xzre · 2026-01-06T16:39:48Z

there seems to be error on build CI because compiler warning, lookin on it

0xzre · 2026-01-06T16:49:01Z

it seems the compiler detects and warns about passing nullptr at compile time, should I use volatile to prevent compiler from optimizing away the nullptr check? @ayushr2
something like:

  volatile void* ptr = nullptr;
  EXPECT_THAT(pwrite64(fd.get(), const_cast<void*>(ptr), 1, 0),
         SyscallFailsWithErrno(EFAULT));

ayushr2 · 2026-01-06T17:52:16Z

I am a bit wary about obfuscating the pointer like that to fool libc. The libc headers declare these functions with attributes (specifically __attribute__((nonnull))) indicating that the buffer pointer must not be null when the size is non-zero.

ayushr2 · 2026-01-06T17:53:58Z

Maybe bypass the libc wrapper using syscall():

// Use syscall() to bypass libc non-null checks.
  EXPECT_THAT(syscall(SYS_pwrite64, fd.get(), nullptr, 1, 0),
              SyscallFailsWithErrno(EFAULT));

// Use syscall() to bypass libc non-null checks.
  EXPECT_THAT(syscall(SYS_write, fd.get(), nullptr, 1),
              SyscallFailsWithErrno(EFAULT));

(be sure to add comments to justify the syscall() usage)

We do this in some other tests as well.

ayushr2 · 2026-01-06T17:56:08Z

Also, could you run clang-tidy on this? Seeing a warning no header providing "SEEK_SET" is directly included in test/syscalls/linux/pwrite64.cc.

0xzre · 2026-01-06T19:15:08Z

Also, could you run clang-tidy on this? Seeing a warning no header providing "SEEK_SET" is directly included in test/syscalls/linux/pwrite64.cc.

hi can help ref which build this warning appear on the CI?
also I ran clang-tidy test/syscalls/linux/pwrite64.cc -- -I. -I./include -I/usr/include -std=c++17 but it the expected warning does not appear 🤔 I did not find also any command to clang-tidy on this codebase

ayushr2 · 2026-01-06T20:43:19Z

Also, could you run clang-tidy on this? Seeing a warning no header providing "SEEK_SET" is directly included in test/syscalls/linux/pwrite64.cc.

hi can help ref which build this warning appear on the CI? also I ran clang-tidy test/syscalls/linux/pwrite64.cc -- -I. -I./include -I/usr/include -std=c++17 but it the expected warning does not appear 🤔 I did not find also any command to clang-tidy on this codebase

This is from our internal CI build. You can probably ignore this as both files already have this warning and it is not introduced by this PR. Lets fix the other issues.

Fixes #2370 Add comprehensive tests for write(2), pwrite(2), and writev(2) syscalls to improve coverage as requested in the issue. New tests include: - Bad buffer states (EFAULT): nullptr buffer, bad iov_base address - Bad file descriptors (EBADF): closed fd, negative fd, read-only fd, O_PATH fd - Offset behavior: verify offset increments, pwrite doesn't change offset - EOF handling: write extends past EOF, pwrite creates holes - Pipes: write to pipe, pwrite returns ESPIPE, write to read end fails - Symlinks: write through symlink, O_NOFOLLOW returns ELOOP - Special files: /dev/null and /dev/zero - writev edge cases: empty iovec, zero-length entries Also adds similar tests to `pwrite64.cc` for consistency. FUTURE_COPYBARA_INTEGRATE_REVIEW=#12426 from 0xzre:write-syscall-tests 6b51a40 PiperOrigin-RevId: 852782177

test/syscalls/linux/pwrite64.cc

Fixes #2370 Add comprehensive tests for write(2), pwrite(2), and writev(2) syscalls to improve coverage as requested in the issue. New tests include: - Bad buffer states (EFAULT): nullptr buffer, bad iov_base address - Bad file descriptors (EBADF): closed fd, negative fd, read-only fd, O_PATH fd - Offset behavior: verify offset increments, pwrite doesn't change offset - EOF handling: write extends past EOF, pwrite creates holes - Pipes: write to pipe, pwrite returns ESPIPE, write to read end fails - Symlinks: write through symlink, O_NOFOLLOW returns ELOOP - Special files: /dev/null and /dev/zero - writev edge cases: empty iovec, zero-length entries Also adds similar tests to `pwrite64.cc` for consistency. FUTURE_COPYBARA_INTEGRATE_REVIEW=#12426 from 0xzre:write-syscall-tests 6b51a40 PiperOrigin-RevId: 852782177

0xzre · 2026-01-07T16:13:05Z

checked the build errors, and found all related on how FUSE translates EFAULT into EIO. skipping the tests that doesn't apply to FUSE fixes this.

local test with bazel test //test/syscalls:write_test_runsc_systrap_fuse //test/syscalls:pwrite64_test_runsc_systrap_fuse attached.

ayushr2 · 2026-01-07T18:27:02Z

found all related on how FUSE translates EFAULT into EIO

I think this might be a bug in our FUSE implementation. Looking.

test/syscalls/linux/write.cc

Fixes #2370 Add comprehensive tests for write(2), pwrite(2), and writev(2) syscalls to improve coverage as requested in the issue. New tests include: - Bad buffer states (EFAULT): nullptr buffer, bad iov_base address - Bad file descriptors (EBADF): closed fd, negative fd, read-only fd, O_PATH fd - Offset behavior: verify offset increments, pwrite doesn't change offset - EOF handling: write extends past EOF, pwrite creates holes - Pipes: write to pipe, pwrite returns ESPIPE, write to read end fails - Symlinks: write through symlink, O_NOFOLLOW returns ELOOP - Special files: /dev/null and /dev/zero - writev edge cases: empty iovec, zero-length entries Also adds similar tests to `pwrite64.cc` for consistency. FUTURE_COPYBARA_INTEGRATE_REVIEW=#12426 from 0xzre:write-syscall-tests 4f4666f PiperOrigin-RevId: 852782177

ayushr2 · 2026-01-08T21:00:13Z

Your tests caught another bug in tmpfs :) This is why testing is important.

I will send a fix for that and then try to get this submitted.

0xzre · 2026-01-08T21:12:29Z

Your tests caught another bug in tmpfs :) This is why testing is important.

I will send a fix for that and then try to get this submitted.

glad the test is proving its worth immediately. let me know if I can help w anything

ayushr2 reviewed Jan 5, 2026

View reviewed changes

test/syscalls/linux/write.cc Outdated Show resolved Hide resolved

test/syscalls/linux/write.cc Outdated Show resolved Hide resolved

test/syscalls/linux/write.cc Outdated Show resolved Hide resolved

ayushr2 reviewed Jan 5, 2026

View reviewed changes

test/syscalls/linux/pwrite64.cc Outdated Show resolved Hide resolved

test/syscalls/linux/pwrite64.cc Outdated Show resolved Hide resolved

0xzre requested a review from ayushr2 January 6, 2026 12:07

0xzre force-pushed the write-syscall-tests branch from 2acf754 to 2b30584 Compare January 6, 2026 15:40

ayushr2 approved these changes Jan 6, 2026

View reviewed changes

ayushr2 added the ready to pull label Jan 6, 2026

copybara-service bot mentioned this pull request Jan 6, 2026

syscalls: beef up write(2) and pwrite(2) tests #12438

Open

0xzre force-pushed the write-syscall-tests branch from 2b30584 to 6b51a40 Compare January 7, 2026 13:58

ayushr2 reviewed Jan 7, 2026

View reviewed changes

test/syscalls/linux/pwrite64.cc Show resolved Hide resolved

0xzre force-pushed the write-syscall-tests branch from 6b51a40 to 7880a4c Compare January 7, 2026 16:13

ayushr2 removed the ready to pull label Jan 7, 2026

ayushr2 reviewed Jan 7, 2026

View reviewed changes

test/syscalls/linux/write.cc Outdated Show resolved Hide resolved

syscalls: Beef up write(2) and pwrite(2) tests

4f4666f

0xzre force-pushed the write-syscall-tests branch from 7880a4c to 4f4666f Compare January 8, 2026 03:46

ayushr2 approved these changes Jan 8, 2026

View reviewed changes

ayushr2 added the ready to pull label Jan 8, 2026

syscalls: beef up write(2) and pwrite(2) tests #12426

Are you sure you want to change the base?

syscalls: beef up write(2) and pwrite(2) tests #12426

Conversation

0xzre commented Jan 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

google-cla bot commented Jan 3, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ayushr2 commented Jan 6, 2026

Uh oh!

0xzre commented Jan 6, 2026

Uh oh!

0xzre commented Jan 6, 2026

Uh oh!

0xzre commented Jan 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ayushr2 commented Jan 6, 2026

Uh oh!

ayushr2 commented Jan 6, 2026

Uh oh!

ayushr2 commented Jan 6, 2026

Uh oh!

0xzre commented Jan 6, 2026

Uh oh!

ayushr2 commented Jan 6, 2026

Uh oh!

Uh oh!

0xzre commented Jan 7, 2026

Uh oh!

ayushr2 commented Jan 7, 2026

Uh oh!

Uh oh!

ayushr2 commented Jan 8, 2026

Uh oh!

0xzre commented Jan 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

0xzre commented Jan 3, 2026 •

edited

Loading

0xzre commented Jan 6, 2026 •

edited

Loading