On Mon, Sep 9, 2013 at 11:43 AM, Stephen Smalley <[email protected]> wrote:
> On 09/09/2013 02:22 PM, Stephen Smalley wrote: > > On 09/09/2013 01:01 PM, Stephen Smalley wrote: > >> On 09/09/2013 09:47 AM, Stephen Smalley wrote: > >>> On 09/06/2013 03:50 PM, Joshua Brindle wrote: > >>>> Add libaudit support for adding directory watch rules. > >>>> > >>>> Add rule parsing support to auditd. > >>>> > >>>> Rule format matches auditctl. Currently only supports -w and -e. > >>>> > >>>> Change-Id: I8bdaea1b5e2a216eec79cd8c9dae583de8295d26 > >>>> > >>>> Signed-off-by: Joshua Brindle <[email protected]> > >>> > >>> Maybe a bug in user, but I did this: > >>> - applied patch and rebuilt, > >>> - reflashed and booted, > >>> - created a /data/misc/audit/audit.rules file that contained: > >>> -w /data/system -p wa > >>> - adb reboot > >>> - adb logcat > logcat.txt > >>> - adb shell su 0 cat /proc/kmsg > dmesg.txt > >>> > >>> logcat.txt showed: > >>> --------- beginning of /dev/log/system > >>> I/auditd ( 119): Starting up > >>> I/audit_log( 119): Previous audit logfile detected, rotating > >>> E/audit_rules( 119): -w /data/system -p wa > >>> > >>> And then nothing else from auditd. > >>> > >>> /data/misc/audit/audit.log has no entries other than the usual: > >>> type=2000 msg=audit(0.710:1): initialized > >>> type=1403 msg=audit(1378733645.695:2): policy loaded auid=4294967295 > >>> ses=4294967295 > >>> type=1404 msg=audit(1378733645.695:3): enforcing=1 old_enforcing=0 > >>> auid=4294967295 ses=4294967295 > >>> type=1403 msg=audit(1378733647.665:4): policy loaded auid=4294967295 > >>> ses=4294967295 > >>> type=1404 msg=audit(1378733830.500:5): enforcing=0 old_enforcing=1 > >>> auid=4294967295 ses=4294967295 > >>> > >>> Creating and deleting files under /data/system appears to do nothing. > >>> What did I miss? > >> > >> So I re-tested with our kernel (i.e. > >> > TARGET_PREBUILT_KERNEL=/path/to/seandroid/kernel/exynos/arch/arm/boot/zImage) > >> and that did generate the expected audit records. I'm guessing that is > >> because we have a patch in our kernel tree that enables audit by > >> default. Since your patch implements the -e (enable) support, I thought > >> I would try that on an unmodified kernel by putting > >> -e 1 > >> -w /data/system -p wa > >> into audit.rules. > >> > >> But we then get a parse error from audit_rules, > >> E/audit_rules( 2504): Could not read audit rules > >> E/auditd ( 2504): error reading audit rules: Try again > >> > >> Am I doing something wrong or is the parser broken? > > > > Actually, it appears that audit_set_enabled() is failing, but your error > > handling code doesn't report it there so it ends up being reported as a > > failure reading the audit rules. > > > > And that in turn appears to be a problem in audit_get_reply(), not > > something you changed. Bill, your code in libaudit.c:audit_get_reply() > > is bailing with an error in the EAGAIN case rather than retrying despite > > the comment to the contrary. > > Also, I think upstream auditd always does an audit_set_enabled(fd, 1) > during startup unless told to do otherwise. > > > With it bailing on EAGAIN, is the mode set to NON_BLOCKING? -- Respectfully, William C Roberts
