From: Adhemerval Zanella <adhemerval.zanella@linaro.org>
To: libc-alpha@sourceware.org,
"Andreas K . Huettel" <dilfridge@gentoo.org>,
Paul Eggert <eggert@cs.ucla.edu>,
Florian Weimer <fweimer@redhat.com>
Subject: [PATCH v6 0/3] Fix opendir regression on some FS
Date: Thu, 2 Mar 2023 11:57:29 -0300 [thread overview]
Message-ID: <20230302145732.2293756-1-adhemerval.zanella@linaro.org> (raw)
Some filesystem might return a non-representable d_off on getdents
call, even if there are few entries in the directories (for instance
ext4, which returns an internal FS hash). This trigger issues with
non-LFS readdir when it finds the first non representable entry, and
also for LFS interface when used along with telldir (which return a
'long int', while d_off is potentially off64_t).
For instance, arm32 dirent and io tests running on qemu-user show the
following failures:
FAIL: dirent/list
FAIL: dirent/tst-scandir
FAIL: io/tst-fts
FAIL: io/tst-fts-lfs
FAIL: io/tst-fts-time64
This patch changes non-LFS readdir so it clamps the d_off value if
it overflows. It now uses getdents64 and maintains an internal map
between d_off that can not be represented by 'long int' (telldir
returned value). seekdir will then set the correct offset by
using the input argument as a key to the internal list.
I have checked on x86_64-linux-gnu, i686-linux-gnu, and
arm-linux-gnueabihf.
Changes from v5:
* Use non-LFS getdents logic on non-LFS readdir.
* Clamp overflow d_off values.
* Removed unrequired readdir64 refactor.
Changes from v4:
* Allocate the telldir map an readdir, thus avoid telldir failure.
* The translation buffer now uses a fixed size struct.
* Fixed the condition to use the long to off64_t map.
Changes from v3:
* Rebase against master.
* Fixed a missing lock unlock on telldir.
* Renamed tst-seekdir2 to tst-opendir-nolfs and check the opendir
result against getdents64.
Adhemerval Zanella (3):
linux: Use getdents64 on non-LFS readdir
support: Add xreallocarray
linux: Set internal DIR filepos as off64_t (BZ #23960, BZ #24050)
include/dirent.h | 2 +-
support/Makefile | 1 +
support/support.h | 2 +
support/xreallocarray.c | 29 ++++
sysdeps/unix/sysv/linux/Makefile | 2 +
sysdeps/unix/sysv/linux/alpha/bits/dirent.h | 3 +
sysdeps/unix/sysv/linux/bits/dirent.h | 4 +
sysdeps/unix/sysv/linux/closedir.c | 4 +
sysdeps/unix/sysv/linux/dirstream.h | 9 +-
sysdeps/unix/sysv/linux/opendir.c | 3 +
sysdeps/unix/sysv/linux/readdir.c | 107 +++++++++++----
sysdeps/unix/sysv/linux/readdir64.c | 11 ++
sysdeps/unix/sysv/linux/rewinddir.c | 5 +
sysdeps/unix/sysv/linux/seekdir.c | 30 +++-
sysdeps/unix/sysv/linux/telldir.c | 36 +++++
sysdeps/unix/sysv/linux/telldir.h | 67 +++++++++
sysdeps/unix/sysv/linux/tst-opendir-lfs.c | 2 +
sysdeps/unix/sysv/linux/tst-opendir.c | 145 ++++++++++++++++++++
18 files changed, 432 insertions(+), 30 deletions(-)
create mode 100644 support/xreallocarray.c
create mode 100644 sysdeps/unix/sysv/linux/telldir.h
create mode 100644 sysdeps/unix/sysv/linux/tst-opendir-lfs.c
create mode 100644 sysdeps/unix/sysv/linux/tst-opendir.c
--
2.34.1
next reply other threads:[~2023-03-02 14:57 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-02 14:57 Adhemerval Zanella [this message]
2023-03-02 14:57 ` [PATCH v6 1/3] linux: Use getdents64 on non-LFS readdir Adhemerval Zanella
2023-03-02 14:57 ` [PATCH v6 2/3] support: Add xreallocarray Adhemerval Zanella
2023-03-10 16:49 ` Florian Weimer
2023-03-10 18:44 ` Adhemerval Zanella Netto
2023-03-10 21:21 ` Paul Eggert
2023-03-02 14:57 ` [PATCH v6 3/3] linux: Set internal DIR filepos as off64_t (BZ #23960, BZ #24050) Adhemerval Zanella
2023-03-10 21:41 ` Paul Eggert
2023-03-13 12:40 ` Adhemerval Zanella Netto
2023-03-13 12:45 ` Adhemerval Zanella Netto
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230302145732.2293756-1-adhemerval.zanella@linaro.org \
--to=adhemerval.zanella@linaro.org \
--cc=dilfridge@gentoo.org \
--cc=eggert@cs.ucla.edu \
--cc=fweimer@redhat.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).