From: Amir Goldstein <amir73il@gmail.com>
To: Arvind Raghavan <raghavan.arvind@gmail.com>
Cc: fstests <fstests@vger.kernel.org>,
Jayashree Mohan <jaya@cs.utexas.edu>,
Vijay Chidambaram <vijay@cs.utexas.edu>
Subject: Re: [PATCH 5/6] src/fssum: Allow single file input
Date: Sun, 31 May 2020 22:31:43 +0300 [thread overview]
Message-ID: <CAOQ4uxjmgtPMsm2MsA=BjgX63-uCKUpp8swHHCkvwUwgLZvMpQ@mail.gmail.com> (raw)
In-Reply-To: <20200531182858.nurku3vcztrzsy43@gmail.com>
On Sun, May 31, 2020 at 9:29 PM Arvind Raghavan
<raghavan.arvind@gmail.com> wrote:
>
> On 05/22, Amir Goldstein wrote:
> > On Fri, May 22, 2020 at 4:06 AM Arvind Raghavan
> > <raghavan.arvind@gmail.com> wrote:
> > >
> > > On 05/21, Amir Goldstein wrote:
> > > > On Thu, May 21, 2020 at 3:10 AM Arvind Raghavan
> > > > <raghavan.arvind@gmail.com> wrote:
> > > > >
> > > > > Allow regular links and symlinks to be passed as input to fssum.
> > > > >
> > > > > Signed-off-by: Arvind Raghavan <raghavan.arvind@gmail.com>
> > > > > Signed-off-by: Jayashree Mohan <jaya@cs.utexas.edu>
> > > > > Signed-off-by: Vijay Chidambaram <vijay@cs.utexas.edu>
> > > > > ---
> > > > > src/fssum.c | 35 ++++++++++++++++++++++++++++++++++-
> > > > > 1 file changed, 34 insertions(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/src/fssum.c b/src/fssum.c
> > > > > index ece0f556..2d1624ca 100644
> > > > > --- a/src/fssum.c
> > > > > +++ b/src/fssum.c
> > > > > @@ -29,6 +29,7 @@
> > > > > #include <inttypes.h>
> > > > > #include <assert.h>
> > > > > #include <endian.h>
> > > > > +#include <libgen.h>
> > > > >
> > > > > #define CS_SIZE 16
> > > > > #define CHUNKS 128
> > > > > @@ -884,8 +885,40 @@ main(int argc, char *argv[])
> > > > > if (gen_manifest)
> > > > > fprintf(out_fp, "Flags: %s\n", flagstring);
> > > > >
> > > > > + struct stat64 path_st;
> > > > > + if (fstat64(fd, &path_st)) {
> > > > > + perror("fstat");
> > > > > + exit(-1);
> > > > > + }
> > > > > +
> > > > > sum_init(&cs);
> > > > > - sum(fd, 1, &cs, path, "");
> > > > > +
> > > > > + if (S_ISDIR(path_st.st_mode)) {
> > > > > + sum(fd, 1, &cs, path, "");
> > > > > + } else if (S_ISREG(path_st.st_mode) || S_ISLNK(path_st.st_mode)) {
> > > > > + // Copy because dirname may modify path
> > > > > + char* path_copy = alloc(strlen(path));
> > > > > + strcpy(path_copy, path);
> >
> > If you stay with this code please use strdup().
> >
> > > > > +
> > > > > + char* dir_path = dirname(path);
> > > > > + char* name = basename(path_copy);
> > > > > +
> > > > > + int dirfd = open(dir_path, O_RDONLY);
> > > > > + if (fd == -1) {
> > > > > + fprintf(stderr, "failed to open %s: %s\n", dir_path,
> > > > > + strerror(errno));
> > > > > + exit(-1);
> > > > > + }
> > > > > +
> > > > > + sum_one(dirfd, 1, &cs, dir_path, "", name);
> > > >
> > > > Instead of all of the above, how about just:
> > > > sum_one(fd, 1, &cs, path, "", "");
> > > >
> > > > From looking at sum_one() code, it seems to me like that will work,
> > > > but I may be missing something.
> > > > It's not that you *want* the name in the checksum, it is not even
> > > > part of the metadata that is being synced with fsync.
> > >
> > > The issue here is that we preserved the code from sum which does
> > > all its opens using openat with the parent directory fd and a
> > > filename. Since we're trying to reuse that code I believe we need
> > > to have this somewhat ugly boilerplate.
> >
> > Ok. But if you stay with this please add a comment about why
> > this is done with a hint for the future how to fix this properly.
> >
> > Or (up to you) you can fix it by calling this helper instead of openat():
> >
> > int open_one(int dirfd, const char *name)
> > {
> > if (!name || !*name)
> > return dup(dirfd);
> > return openat(dirfd, name, 0);
> > }
> >
> > fstatat() can take empty name with AT_EMPTY_PATH flag.
> > readlinkat() should be able to take an empty name, but documentation
> > is not clear whether fd must be O_PATH - need to verify if it works with
> > non O_PATH fd.
>
> I might be doing something wrong but I can't seem to get
> AT_EMPTY_PATH to work with fstatat. I don't see it on the man
> page and when I pass it to fstatat it errors out with invalid
> argument.
>
https://www.man7.org/linux/man-pages/man2/stat.2.html
Says:
AT_EMPTY_PATH (since Linux 2.6.39)
... This flag is Linux-specific; define _GNU_SOURCE to obtain its definition.
And I see that fssum.c does define _GNU_SOURCE, so not sure what the
problem is.
> In the case that fstatat doesn't support AT_EMPTY_PATH, do you
> see another way to fix the above code?
>
Yes, you can pass path_st to sum_one() you already have it.
But I don't think that you or anyone that cares about xfstests is running
a kernel older than 2.6.39...
Thanks,
Amir.
next prev parent reply other threads:[~2020-05-31 19:31 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-20 21:16 [PATCH 0/6] Changes to fssum to support POSIX Arvind Raghavan
2020-05-20 21:18 ` [PATCH 1/6] src/fssum: Make sum_file_data global Arvind Raghavan
2020-05-20 21:19 ` [PATCH 2/6] src/fssum: Refactor recursive traversal Arvind Raghavan
2020-05-20 21:19 ` [PATCH 3/6] src/fssum: Add flag -R for non-recursive mode Arvind Raghavan
2020-05-20 21:20 ` [PATCH 4/6] src/fssum: Add a flag for including file size in checksum Arvind Raghavan
2020-05-20 21:21 ` [PATCH 5/6] src/fssum: Allow single file input Arvind Raghavan
2020-05-21 9:18 ` Amir Goldstein
2020-05-22 1:06 ` Arvind Raghavan
2020-05-22 5:37 ` Amir Goldstein
2020-05-31 18:28 ` Arvind Raghavan
2020-05-31 19:31 ` Amir Goldstein [this message]
2020-06-21 23:07 ` Arvind Raghavan
2020-05-20 21:21 ` [PATCH 6/6] src/fssum: Fix whitespace in usage Arvind Raghavan
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='CAOQ4uxjmgtPMsm2MsA=BjgX63-uCKUpp8swHHCkvwUwgLZvMpQ@mail.gmail.com' \
--to=amir73il@gmail.com \
--cc=fstests@vger.kernel.org \
--cc=jaya@cs.utexas.edu \
--cc=raghavan.arvind@gmail.com \
--cc=vijay@cs.utexas.edu \
/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 NNTP newsgroup(s).