git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v3 2/3] sha1_file: open window into packfiles with O_CLOEXEC

From
Linus Torvalds <torvalds@linux-foundation.org>
Date
Oct 28, 2016, 17:38 UTC
Message-ID
<CA+55aFwUEzfvWVSZfhBi85QaKWSo-gVMOk1BJFrR0ZsdCRHRsg@mail.gmail.com>
In-Reply-To
<xmqqshrg1ksv.fsf@gitster.mtv.corp.google.com>
On Fri, Oct 28, 2016 at 9:48 AM, Junio C Hamano <gitster@pobox.com> wrote:
>
> Excuse me, but care to elaborate a bit more?
It just seems entirely pointless to change the games with O_xyz.
> Setting FD_CLOEXEC with fcntl(2) may be racy, but in what way
> it is broken to open(2) with O_CLOEXEC?

Apparently windows doesn't even support it, at least not mingw. So you'll have to have some supprt for just explicitly closing things at execve anyway. And if you do that for windows, then what's the point of using O_CLOEXEC on Linux, and having two totally different paths that will just mean maintenance headache.

In contrast, O_NOATIME isn't a maintenance problem, since it's purely an optimization and has no semantic difference, so it not existing on some platform is immaterial.

End result: leave O_NOATIME as it is (working), and just forget about O_CLOEXEC.

I have no actual idea about mingw support for this, but Johannes' email certainly implies it's painful on Windows. Some googling also shows people doing things like

  #ifdef O_CLOEXEC
          flags |= O_CLOEXEC;
  #endif
or
  #ifndef O_CLOEXEC
  #define O_CLOEXEC 0
  #endif

so O_CLOEXEC definitely isn't considered portable. Do you really want to rely on it?

              Linus
Previous: Junio C HamanoNext: Junio C Hamano
Message 37 of 54 in “Use CLOEXEC to avoid fd leaks”
  1. 0/2 Use CLOEXEC to avoid fd leakslarsxschneider@gmail.com, Oct 24, 2016
  2. 1/2 sha1_file: open window into packfiles with CLOEXEClarsxschneider@gmail.com, Oct 24, 2016
  3. Johannes SchindelinOct 25, 2016
  4. Junio C HamanoOct 25, 2016
  5. 2/2 read-cache: make sure file handles are not inherited by child processeslarsxschneider@gmail.com, Oct 24, 2016
  6. Eric WongOct 24, 2016
  7. Junio C HamanoOct 24, 2016
  8. Johannes SchindelinOct 25, 2016
  9. Junio C HamanoOct 25, 2016
  10. Johannes SixtOct 24, 2016
  11. Lars SchneiderOct 24, 2016
  12. Johannes SixtOct 25, 2016
  13. Junio C HamanoOct 24, 2016
  14. Johannes SchindelinOct 25, 2016
  15. 0/3 quick reroll of Lars's git_open() w/ O_CLOEXECJunio C Hamano, Oct 25, 2016
  16. 2/3 sha1_file: open window into packfiles with O_CLOEXECJunio C Hamano, Oct 25, 2016
  17. Jeff KingOct 26, 2016
  18. Junio C HamanoOct 26, 2016
  19. Jeff KingOct 26, 2016
  20. Junio C HamanoOct 26, 2016
  21. Jeff KingOct 26, 2016
  22. Junio C HamanoOct 26, 2016
  23. Jeff KingOct 27, 2016
  24. Junio C HamanoOct 27, 2016
  25. Linus TorvaldsOct 27, 2016
  26. Junio C HamanoOct 27, 2016
  27. Linus TorvaldsOct 27, 2016
  28. Linus TorvaldsOct 27, 2016
  29. Junio C HamanoOct 27, 2016
  30. Linus TorvaldsOct 27, 2016
  31. Junio C HamanoOct 28, 2016
  32. Junio C HamanoOct 28, 2016
  33. Eric WongOct 28, 2016
  34. Johannes SchindelinOct 28, 2016
  35. Linus TorvaldsOct 28, 2016
  36. Junio C HamanoOct 28, 2016
  37. Linus TorvaldsOct 28, 2016
  38. Junio C HamanoOct 28, 2016
  39. Junio C HamanoOct 29, 2016
  40. Johannes SchindelinOct 29, 2016
  41. Linus TorvaldsOct 29, 2016
  42. Junio C HamanoOct 31, 2016
  43. Jeff KingOct 31, 2016
  44. Junio C HamanoOct 31, 2016
  45. Jeff KingOct 31, 2016
  46. Junio C HamanoOct 28, 2016
  47. Junio C HamanoOct 28, 2016
  48. Jeff KingOct 28, 2016
  49. 1/3 sha1_file: rename git_open_noatime() to git_open()Junio C Hamano, Oct 25, 2016
  50. 3/3 read-cache: make sure file handles are not inherited by child processesJunio C Hamano, Oct 25, 2016
  51. Eric WongOct 25, 2016
  52. Junio C HamanoOct 25, 2016
  53. Lars SchneiderOct 25, 2016
  54. Junio C HamanoOct 25, 2016

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.