[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <878rvj1tut.fsf@alyssa.is>
Date: Thu, 13 Jan 2022 16:30:18 +0000
From: Alyssa Ross <hi@...ssa.is>
To: Rich Felker <dalias@...c.org>
Cc: musl@...ts.openwall.com
Subject: Re: [PATCH musl v2 3/3] mntent: fix parsing lines with
optional fields
Hi Rich, thanks for following up on this.
Rich Felker <dalias@...c.org> writes:
> On Mon, Sep 20, 2021 at 12:21:41AM -0400, Rich Felker wrote:
>> On Wed, Sep 15, 2021 at 10:11:55PM +0000, Alyssa Ross wrote:
>> > According to fstab(5), the last two fields are optional, but this
>> > wasn't accepted by Musl. After this change, only the first field is
>> > required, which matches Glibc's behaviour.
>> >
>> > Using sscanf as before, it would have been impossible to differentiate
>> > between 0 fields and 4 fields, because sscanf would have returned 0 in
>> > both cases due to the use of assignment suppression and %n for the
>> > string fields (which is important to avoid copying any strings). So
>> > instead, before calling sscanf, initialize every string to the empty
>> > string, and then we can check which strings are empty afterwards to
>> > know how many fields were matched.
>> > ---
>> >
>> > We could also be stricter about it, and enforce that the first four
>> > fields are present, since the man page says only the last two are
>> > optional. Doing that would be a simple change of checking for the
>> > presence of mnt_opts instead of mnt_fsname at the end of my patch.
>> >
>> > v2: don't change n from int to size_t
>> >
>> > src/misc/mntent.c | 18 +++++++++++++-----
>> > 1 file changed, 13 insertions(+), 5 deletions(-)
>> >
>> > diff --git a/src/misc/mntent.c b/src/misc/mntent.c
>> > index eabb8200..238a0efd 100644
>> > --- a/src/misc/mntent.c
>> > +++ b/src/misc/mntent.c
>> > @@ -21,7 +21,8 @@ int endmntent(FILE *f)
>> >
>> > struct mntent *getmntent_r(FILE *f, struct mntent *mnt, char *linebuf, int buflen)
>> > {
>> > - int cnt, n[8], use_internal = (linebuf == SENTINEL);
>> > + int n[8], use_internal = (linebuf == SENTINEL);
>> > + size_t len, i;
>> >
>> > mnt->mnt_freq = 0;
>> > mnt->mnt_passno = 0;
>> > @@ -39,10 +40,14 @@ struct mntent *getmntent_r(FILE *f, struct mntent *mnt, char *linebuf, int bufle
>> > errno = ERANGE;
>> > return 0;
>> > }
>> > - cnt = sscanf(linebuf, " %n%*s%n %n%*s%n %n%*s%n %n%*s%n %d %d",
>> > - n, n+1, n+2, n+3, n+4, n+5, n+6, n+7,
>> > - &mnt->mnt_freq, &mnt->mnt_passno);
>> > - } while (cnt < 2 || linebuf[n[0]] == '#');
>> > +
>> > + len = strlen(linebuf);
>> > + for (i = 0; i < sizeof n / sizeof *n; i++) n[i] = len;
>> > + if (sscanf(linebuf, " %n%*s%n %n%*s%n %n%*s%n %n%*s%n %d %d",
>> > + n, n+1, n+2, n+3, n+4, n+5, n+6, n+7,
>> > + &mnt->mnt_freq, &mnt->mnt_passno) == EOF && ferror(f))
>> > + return 0;
>> > + } while (linebuf[n[0]] == '#');
>> >
>> > linebuf[n[1]] = 0;
>> > linebuf[n[3]] = 0;
>> > @@ -54,6 +60,9 @@ struct mntent *getmntent_r(FILE *f, struct mntent *mnt, char *linebuf, int bufle
>> > mnt->mnt_type = linebuf+n[4];
>> > mnt->mnt_opts = linebuf+n[6];
>> >
>> > + if (!*mnt->mnt_fsname)
>> > + return 0;
>> > +
>> > return mnt;
>> > }
>>
>> It looks like your patch changes the behavior for malformed lines from
>> skipping them (and continuing to search for the next valid line) to
>> returning 0. Is that intentional? Maybe it's better; I'm not sure. But
>> won't it even cause blank lines to return 0?
Because I only check for the first field being present, a lot of
nonsensical lines will be accepted.
As I said in the patch commentary:
> After this change, only the first field is
> required, which matches Glibc's behaviour.
>
> [snip]
>
> We could also be stricter about it, and enforce that the first four
> fields are present, since the man page says only the last two are
> optional. Doing that would be a simple change of checking for the
> presence of mnt_opts instead of mnt_fsname at the end of my patch.
It probably would make more sense to check that the four fields the man
pages implies are required are all there, by making the change I
suggested in the commentary, at least until somebody complains about
their two-field fstab being accepted by Glibc and not Musl.
> Indeed it also seems to be skipping empty lines, contrary to what you
> said in another message:
>
>> • Empty lines should be skipped.
Yes, it looks like I was mistaken before when I thought that Musl didn't
properly handle comments and empty lines. Looking back, it seems that
the tests I was running were against mntent files with only four fields,
so the parsing failures I was seeing were because of that, not because
of an issue in the comment or empty line handling.
My patch does (inadventently) change the behaviour of empty line
handling. We should leave the current behaviour of skipping over empty
lines as is. Suggested fix at the end of this message.
> Do you have a preference on how to proceed? We could add back a
> condition to the while loop, something like linebuf[n[0]]=='#' ||
> n[6]==len (i.e. skip lines with too few fields, possibly using a
> different number instead of 6 if more appropriate). Or we could do
> what I suggested before:
>
>> A less invasive change might be adding "%1[ \t\n\v\f\r]" and a dummy
>> char* argument to collect the value before the " %d %d". Then you can
>> check for cnt<1. But I'm not sure even the 4th field should be
>> mandatory. This same apprach could be used to make just 3 mandatory if
>> desired though.
>
> Thoughts?
I think it would clearer to have an explicit check that the last
mandatory field is set, like I currently do with the mnt_fsname check at
the end of the function. I don't particularly mind how many fields are
mandatory, as long as its four or fewer so Musl's behaviour follows the
fstab format described in the man page.
So overall my proposed revisions would be the following. There's a
change to move to the next line if the current one is empty, and a
change to ensure the first four fields are all present. (If you decide
you'd like the fourth field to also be optional, we can just change
mnt_opts to mnt_type in that check.) It's been a long time since I last
this code, btw, so I hope I'm not missing anything around the empty line
check. I'd be happy to put together a revised series, with these
changes and a corresponding change to my libc-test patch, if you'd like.
diff --git i/src/misc/mntent.c w/src/misc/mntent.c
index 169e9789..7782cb10 100644
--- i/src/misc/mntent.c
+++ w/src/misc/mntent.c
@@ -47,7 +47,7 @@ struct mntent *getmntent_r(FILE *f, struct mntent *mnt, char *linebuf, int bufle
n, n+1, n+2, n+3, n+4, n+5, n+6, n+7,
&mnt->mnt_freq, &mnt->mnt_passno) == EOF && ferror(f))
return 0;
- } while (linebuf[n[0]] == '#');
+ } while (linebuf[0] == '\n' || linebuf[n[0]] == '#');
linebuf[n[1]] = 0;
linebuf[n[3]] = 0;
@@ -59,7 +59,7 @@ struct mntent *getmntent_r(FILE *f, struct mntent *mnt, char *linebuf, int bufle
mnt->mnt_type = linebuf+n[4];
mnt->mnt_opts = linebuf+n[6];
- if (!*mnt->mnt_fsname)
+ if (!*mnt->mnt_opts)
return 0;
return mnt;
Download attachment "signature.asc" of type "application/pgp-signature" (833 bytes)
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
