If the size of the error log is too big to send via email, and the sending
fails, it wont email any result. This can be confusing for the user who is
waiting for an email on the completion of the tests.

If it fails to send email, then try again without the log file stating that
it failed to send an email. Obviously this will not be of use if the sending
of email failed for some other reasons, but it will at least give the user
some information when it fails for the most common reason.

Cc: stable@vger.kernel.org
Fixes: c2d84ddb ("ktest.pl: Add MAIL_COMMAND option to define how to send email")
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

This issue was first noticed when I was testing different kernels on
Oracle Linux 8 which as Fedora 30+ adopts BLS as default. Even though a
kernel entry was added successfully and the index of that kernel entry was
retrieved correctly, ktest still wouldn't reboot the system into
user-specified kernel.

The bug was spotted in subroutine reboot_to where the if-statement never
checks for REBOOT_TYPE "grub2bls", therefore the desired entry will not be
set for the next boot.

Add a check for "grub2bls" so that $grub_reboot $grub_number can
be run before a reboot if REBOOT_TYPE is "grub2bls" then we can boot to
the correct kernel.

Link: https://lkml.kernel.org/r/20201121021243.1532477-1-libo.chen@oracle.com

Cc: stable@vger.kernel.org
Fixes: ac246645 ("ktest: introduce grub2bls REBOOT_TYPE option")
Signed-off-by: NLibo Chen <libo.chen@oracle.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

There is a spelling mistake in an error message. Fix it.

Link: https://lkml.kernel.org/r/20200810100750.61475-1-colin.king@canonical.comSigned-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If the log file for a given test is larger than the max size given then use
set the seek from the end of the log file instead of from the start of the
test.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Add the ktest config option MAIL_MAX_SIZE that will limit the size of the
log file that is placed into the email on failure.

Link: https://lore.kernel.org/r/20200701231756.790637968@goodmis.orgReviewed-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If a failure happens and an email is sent, show the contents of the log of
the last test that failed in the email.

Link: http://lore.kernel.org/r/20200701231756.619246244@goodmis.orgReviewed-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

The log file should be up to date to whatever is happening in ktest.
Disable buffering to the LOG output file handle.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Currently, every write to the log file is done by opening the file, writing
to it, then closing the file. This rather expensive. Just open it at the
beginning and close it at the end.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

There is a NOT DEFINED operator, but there is not an operator that can
negate any other expression.

 For example: NOT (${FOO} == boot || ${BAR} == run)

Add the keyword NOT to allow the ktest.pl config files to negate operators.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Currently, if a PRE_TEST is defined and ran, but fails, there's nothing
currently available to make the test fail too. Add a PRE_TEST_DIE option that
when set, if a PRE_TEST is defined and fails, the test will die too.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If a log file is defined and the test were to error, a print statement is
made that shows the user where the log file is to examine it further. But
this is not done if the test were to succeed.

I find it annoying that it does not show where the log file is on success,
as I run several different tests that place their log files in various
locations, and even though the test pass, there's things I want to look at
in the log file (like warnings). It is much easier to find where the log
file is, if it is displayed at the end of a test.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

When performing a automatic config bisect via ktest.pl, it is very useful to
have a copy of each of the bisects used. This way, if a bisect were to go
wrong, it is possible to retrace the steps and continue at the location
before the error was made.

The ktest.pl will make a copy of the good and bad configs, labeled as such,
as well as a number attached to it that represents the iteration of the
bisect. These files are saved in the ktest temp directory where it currently
stores the good and bad config files.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

This patch fixes multipe spelling typo found in ktest.pl.

Link: http://lkml.kernel.org/r/20200309115430.57540-1-standby24x7@gmail.comAcked-by: NRandy Dunlap <rdunlap@infradead.org>
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Before rebooting the box, a "ssh sync" is called to the test machine to see
if it is alive or not. But if the test machine is in a partial state, that
ssh may never actually finish, and the ktest test hangs.

Add a 10 second timeout to the sync test, which will fail after 10 seconds
and then cause the test to reboot the test machine.

Cc: stable@vger.kernel.org
Fixes: 6474ace9 ("ktest.pl: Powercycle the box on reboot if no connection can be made")
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

For the last time, I screwed up my ktest config file, and the build went
into the default "randconfig", blowing away the .config that I had set up.
The reason for the default randconfig was because when this was first
written, I wanted to do a bunch of randconfigs. But as time progressed,
ktest isn't about randconfig anymore, and because randconfig destroys the
config in the build directory, it's a dangerous default to have. Use
oldconfig as the default.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Based on 1 normalized pattern(s):

  licensed under the terms of the gnu gpl license version 2

extracted by the scancode license scanner the SPDX license identifier

  GPL-2.0-only

has been chosen to replace the boilerplate/reference in 62 file(s).
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NAllison Randal <allison@lohutok.net>
Reviewed-by: NKate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: NRichard Fontana <rfontana@redhat.com>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070033.929121379@linutronix.deSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove get_grub2_index() because it isn't used anywhere.

Link: http://lkml.kernel.org/r/20190509213647.6276-6-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

For BLS, kernel entry is added by kernel-install command through
POST_INSALL, for example,

POST_INSTALL = ssh root@Test "/usr/bin/kernel-install \
    add $KERNEL_VERSION /boot/vmlinuz-$KERNEL_VERSION"

The entry is removed by kernel-install command and the kernel
version is needed for the argument.

Pass KERNEL_VERSION variable to POST_KTEST so that kernel-install
command can remove the entry like as follows:

POST_KTEST = ssh root@Test "/usr/bin/kernel-install remove $KERNEL_VERSION"

Link: http://lkml.kernel.org/r/20190509213647.6276-5-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Fedora 30 introduces Boot Loader Specification (BLS),
it changes around grub entry configuration.

kernel entries aren't in grub.cfg. We can get the entries
by "grubby --info=ALL" command.

Introduce grub2bls as REBOOT_TYPE option for BLS.

Link: http://lkml.kernel.org/r/20190509213647.6276-4-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Cleanup get_grub_index().

Link: http://lkml.kernel.org/r/20190509213647.6276-3-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Introduce _get_grub_index() to deal with Boot Loader
Specification (BLS) and cleanup.

Link: http://lkml.kernel.org/r/20190509213647.6276-2-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Unexpected power cycle occurs while the installation of the
kernel.

   ssh root@Test sync ... [0 seconds] SUCCESS
   ssh root@Test reboot ... [1 second] FAILED!
   virsh destroy Test; sleep 5; virsh start Test ... [6 seconds] SUCCESS

That is because REBOOT, the default is "ssh $SSH_USER@$MACHINE
reboot", exits as 255 even if the reboot is successfully done,
like as:

   ]# ssh root@Test reboot
   Connection to Test closed by remote host.
   ]# echo $?
   255
   ]#

To avoid the unexpected power cycle, introduce a new parameter,
REBOOT_RETURN_CODE to judge whether REBOOT is successfully done
or not.

Link: http://lkml.kernel.org/r/20190418135943.12640-1-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

ktest fails if meta characters are in GRUB_MENU, for example
GRUB_MENU = 'Fedora (test)'

The failure happens because the meta characters are not escaped,
so the menu doesn't match in any entries in GRUB_FILE.

Use quotemeta() to escape the meta characters.

Link: http://lkml.kernel.org/r/20190417235823.18176-1-msys.mizuma@gmail.comSigned-off-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If a test has an error, display not only the what type of test failed, but
if the test was giving a name, display that too, as well as the current
iteration of the tests. Each test has an iteration number associated to it.
For error messages display that iteration number along with the test type
and test name. This includes the message that gets sent via email.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

grub-reboot selects the submenu's first menuentry (title is "1>0") rather than ktest's
menuentry (title is "2") by mistake.

===
$ sudo cat /boot/grub/grub.cfg  | grep -E "^menuentry|^submenu"
...
menuentry 'Ubuntu' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option '...' {
...
submenu 'Advanced options for Ubuntu' $menuentry_id_option '...' {
...
menuentry 'ktest' {
...
===

Correct it by taking submenu entries into account in get_grub2_index().

Link: http://lkml.kernel.org/r/87poaje4as.wl-satoru.takeuchi@gmail.comSigned-off-by: NSatoru Takeuchi <satoru.takeuchi@gmail.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Allow the user to override the default way to send email. This will allow
the user to add their own mailer and format for sending email.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Instead of open coding system() call, use run_command which will log the
sending of email as well.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If dodie cause a function that itself will call dodie, then be able to
handle that. This will allow dodie functions to call run_command, which
could possibly call dodie. If dodie is called again, simply ignore it.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If the user specifies a MAILTO, but the MAILER is not supported, then
kill the test.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

The option MAIL_PATH lets the user decide how to find the mailer they are
using. For example, sendmail is usually located in /usr/sbin but is not
always in the path of non admin users. Have ktest look through the user's
PATH environment variable (adding /usr/sbin) as well, but if that's not good
enough, allow the user to define where to find the mailer.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

squash to mail exec
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If the user doesn't want to send mail, then don't bother them with output
that says they didn't specify a mailer. That can be annoying.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Users should get emails when the script dies because of a critical failure.
Critical failures are defined as any errors that could abnormally terminate
the script.

In order to add email support, this patch converts all die() to dodie() except:
 * when '-v' is used as an option to get the version of the script.
 * in Sig-Int handeler because it's not a fatal error to cancel the script.
 * errors happen during parsing config

Link: http://lkml.kernel.org/r/1522094884-22718-4-git-send-email-tianyang.chen@oracle.comSuggested-by: NDhaval Giani <dhaval.giani@oracle.com>
Signed-off-by: NTim Tianyang Chen <tianyang.chen@oracle.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

User can cancel tests and specify handler's behavior using option
'EMAIL_WHEN_CANCELED'.

Link: http://lkml.kernel.org/r/1522094884-22718-3-git-send-email-tianyang.chen@oracle.comSuggested-by: NDhaval Giani <dhaval.giani@oracle.com>
Signed-off-by: NTim Tianyang Chen <tianyang.chen@oracle.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Users can define optional variables to get email notifications.
Ktest can send emails when the script:
 * was started
 * failed with fatal errors and called dodie()
 * completed all testing

Users have to setup the mailer provided in config prior to using this script.
Supported mailers: mailx, mail, sendmail
mailer specific routines are _sendmail_send(), _mailx_send()

Link: http://lkml.kernel.org/r/1522094884-22718-2-git-send-email-tianyang.chen@oracle.comSuggested-by: NDhaval Giani <dhaval.giani@oracle.com>
Signed-off-by: NTim Tianyang Chen <tianyang.chen@oracle.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If a config-bisect was interrupted, then allow the user to continue, or
restart a new config-bisect.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Just looking for config-bisect.pl in the source tree can be risky,
especially, if the source tree being tested doesn't have config-bisect.pl in
place. Instead, allow the user to set where to find config-bisect.pl with a
new option CONFIG_BISECT_EXEC.

If this option is not set, by default, ktest.pl will look for
config-bisect.pl in the following locations:

 `pwd`/config-bisect.pl # where ktest.pl was called from
 `dirname /path/to/ktest.pl`/config-bisect.pl # where ktest.pl exists
 ${BUILD_DIR}/tools/testing/ktest/config-bisect.pl
   # where config-bisect.pl exists in the source tree.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If config-bisect.pl sees that a config_bisect has already been started, it
will ask on the command line if it should bisect or not. This will mess up
running config_bisect from ktest.pl.

Have ktest.pl pass in '-r' to config-bisect.pl and have config-bisect.pl
recognize that to reset without asking.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Check to see if diffconfig is available and use that to diff the configs
instead of using 'diff -u', as diffconfig produces much better output of
kernel config files. It checks the source directory for the executable.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

When commands are run in ktest, they are only displayed in the ktest log
file, but that is not sufficient for outputting the display for config
bisects. The result of a config bisect is not shown.

Add a way to display the output of "run_command" which is the subroutine
used by ktest to execute commands. Use this feature to display the output of
config-bisect.pl executions to see the progress as well as the result.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Reduce code duplication and take advantage of bisection logic
improvements by calling config-bisect.pl.

The output of make oldconfig is now copied directly to the desired file,
rather than doing assign_configs+save_config, in order to preserve the
ordering so that diffing the configs at the end will provide useful
output.

Link: http://lkml.kernel.org/r/20170717001630.10518-8-swood@redhat.comSigned-off-by: NScott Wood <swood@redhat.com>
[ Modified to use with new version of config-bisect.pl ]
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>