public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Guillem Jover <guillem@debian.org>
To: Florian Weimer <fweimer@redhat.com>
Cc: libc-alpha@sourceware.org
Subject: Re: [PATCH] ldconfig: Actually ignore dpkg-related temporarily files
Date: Tue, 24 Oct 2023 13:45:21 +0200	[thread overview]
Message-ID: <ZTeuUUZATbMpiU0B@thunder.hadrons.org> (raw)
In-Reply-To: <87edhlwq6y.fsf@oldenburg.str.redhat.com>

[-- Attachment #1: Type: text/plain, Size: 2378 bytes --]

Hi!

On Mon, 2023-10-23 at 13:09:25 +0200, Florian Weimer wrote:
> These files actually end in .dpkg-new and .dpkg-tmp, not .tmp
> as I mistakenly assumed.
> 
> Fixes commit 2aa0974d2573441bffd59 ("elf: ldconfig should skip
> temporary files created by package managers").
> 
> ---
>  elf/ldconfig.c | 17 ++++++++++++++---
>  1 file changed, 14 insertions(+), 3 deletions(-)
> 
> diff --git a/elf/ldconfig.c b/elf/ldconfig.c
> index 02387a169c..4abde26e60 100644
> --- a/elf/ldconfig.c
> +++ b/elf/ldconfig.c
> @@ -661,6 +661,17 @@ struct dlib_entry
>    struct dlib_entry *next;
>  };
>  
> +/* Return true if the N bytes at NAME end with with the characters in
> +   the string SUFFIX.  (NAME[N + 1] does not have to be a null byte.)
> +   Expected to be called with a string literal for SUFFIX.  */
> +static inline bool
> +endswithn (const char *name, size_t n, const char *suffix)
> +{
> +  return (n >= strlen (suffix)
> +	  && memcmp (name + n - strlen (suffix), suffix,
> +		     strlen (suffix)) == 0);
> +}
> +
>  /* Skip some temporary DSO files.  These files may be partially written
>     and lead to ldconfig crashes when examined.  */
>  static bool
> @@ -670,8 +681,7 @@ skip_dso_based_on_name (const char *name, size_t len)
>       names like these are never really DSOs we want to look at.  */
>    if (len >= sizeof (".#prelink#") - 1)
>      {
> -      if (strcmp (name + len - sizeof (".#prelink#") + 1,
> -		  ".#prelink#") == 0)
> +      if (endswithn (name, len, ".#prelink#"))
>  	return true;
>        if (len >= sizeof (".#prelink#.XXXXXX") - 1
>  	  && memcmp (name + len - sizeof (".#prelink#.XXXXXX")
> @@ -682,7 +692,8 @@ skip_dso_based_on_name (const char *name, size_t len)
>    if (memchr (name, len, ';') != NULL)
>      return true;
>    /* Skip temporary files created by dpkg.  */
> -  if (len > 4 && memcmp (name + len - 4, ".tmp", 4) == 0)
> +  if (endswithn (name, len, ".dpkg-new")
> +      || endswithn (name, len, ".dpkg-tmp"))
>      return true;
>    return false;
>  }
> 
> base-commit: 2aa0974d2573441bffd596b07bff8698b1f2f18c

Thanks, LGTM!

I've tentatively queued the attached patch (will review it further for
correctness later today) trying to document these files in dpkg(1), as
I noticed this does not seem to be documented anywhere (except in code
comments), in the hopes it will help the next person. :)

Thanks,
Guillem

[-- Attachment #2: 0001-man-Document-the-intermediary-files-used-during-unpa.patch --]
[-- Type: text/x-diff, Size: 2490 bytes --]

From 6e30d7f80b32a09511c1b849e392f87768ddcaef Mon Sep 17 00:00:00 2001
From: Guillem Jover <guillem@debian.org>
Date: Sun, 22 Oct 2023 12:50:49 +0200
Subject: [PATCH] man: Document the intermediary files used during unpack and
 configure

There did not seem to be any explicit documentation anywhere about these
files nor their functions.

Prompted-by: Florian Weimer <fweimer@redhat.com>
Ref: https://sourceware.org/pipermail/libc-alpha/2023-October/152245.html
---
 man/dpkg.pod | 45 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/man/dpkg.pod b/man/dpkg.pod
index 8594e8e71..78383c50d 100644
--- a/man/dpkg.pod
+++ b/man/dpkg.pod
@@ -1448,6 +1448,51 @@ useful if it's lost or corrupted due to filesystems troubles.
 
 The format and contents of a binary package are described in L<deb(5)>.
 
+=head2 Filesystem files
+
+During unpacking and configuration B<dpkg> uses various files for backup
+and rollback purposes.
+
+=over
+
+=item F<*.dpkg-new>
+
+During unpack, B<dpkg> extracts all new filesystem objects
+into I<filename>B<.dpkg-new>
+(except for new directories which go directly into I<filename>,
+or existing directories or symlinks to directories which get skipped),
+once that is done and after having performed backups of the old objects,
+eventually all get renamed to I<filename>.
+
+=item F<*.dpkg-tmp>
+
+During unpack, B<dpkg> makes backups of the old filesystem objects
+into I<filename>B<.dpkg-tmp> after extracting the new files.
+These backups are not performed for conffiles as they are processed
+at a later stage.
+Otherwise these backups are performed as either a rename for directories
+(but only if they switch file type),
+a new symlink copy for symlinks,
+or a hard link for any other filesystem object.
+
+In case of needing to rollback, these backups get used to restore the
+previous contents of the objects.
+These get removed automatically after the installation is complete.
+
+=item F<*.dpkg-old>
+
+During configuration, when installing a new version,
+B<dpkg> cam make a backup of the previous modified conffile into
+I<filename>B<.dpkg-old>.
+
+=item F<*.dpkg-dist>
+
+During configuration, when keeping the old version,
+B<dpkg> can make a backup of the new unmodified conffile into
+I<filename>B<.dpkg-dist>.
+
+=back
+
 =head1 SECURITY
 
 Any operation that needs write access to the database or the filesystem
-- 
2.42.0


  reply	other threads:[~2023-10-24 12:09 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-23 11:09 Florian Weimer
2023-10-24 11:45 ` Guillem Jover [this message]
2024-01-08 21:17   ` Guillem Jover

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=ZTeuUUZATbMpiU0B@thunder.hadrons.org \
    --to=guillem@debian.org \
    --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).