| Message ID | 20190903210537.29142-1-ira.weiny@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Add lease testing | expand |
On Tue, Sep 03, 2019 at 02:05:35PM -0700, ira.weiny@intel.com wrote: > From: Ira Weiny <ira.weiny@intel.com> > > To get to a point of developing and testing the new Lease semantics for long > term file pins[1], add a simple lease test. Hi Ira, Can you co-ordinate with Jeff on this? He just posted a patch to add F_SETLEASE testing with a new binary a few days ago... https://marc.info/?l=fstests&m=156690934513424&w=2 Cheers, Dave.
On Wed, Sep 04, 2019 at 07:42:55AM +1000, Dave Chinner wrote: > On Tue, Sep 03, 2019 at 02:05:35PM -0700, ira.weiny@intel.com wrote: > > From: Ira Weiny <ira.weiny@intel.com> > > > > To get to a point of developing and testing the new Lease semantics for long > > term file pins[1], add a simple lease test. > > Hi Ira, > > Can you co-ordinate with Jeff on this? He just posted a patch > to add F_SETLEASE testing with a new binary a few days > ago... > > https://marc.info/?l=fstests&m=156690934513424&w=2 Sure. Jeff, Looking at your submission revealed I was missing 2 tests (shown in the diff below). Would it be sufficient to add these to this submission and go with my series? Or would you prefer using fork vs the socket harness? I think we may need a fork in the server side of my tests to work on duplicated FD's but right now I have more tests than you. Let me know, Thanks, Ira commit 849633871b7c0b00a6ee2bb751c09cd6365da386 (lease-test-b0-v3) Author: Ira Weiny <ira.weiny@intel.com> Date: Tue Sep 3 15:17:52 2019 -0700 src/leasetest: Add tests to match competing test A competing patch to test leases was submitted in parallel.[1] This adds tests which were covered in that patch set. [1] https://marc.info/?l=fstests&m=156690934513424&w=2 Signed-off-by: Ira Weiny <ira.weiny@intel.com> diff --git a/src/leasetest.c b/src/leasetest.c index b30a4d05d039..729aca851493 100644 --- a/src/leasetest.c +++ b/src/leasetest.c @@ -137,6 +137,8 @@ char *descriptions[] = { /* 4 */"Fail Read Lease if opened with write permissions", /* 5 */"Read lease gets SIGIO on write open", /* 6 */"Write lease gets SIGIO on read open", + /* 7 */"No SIGIO is sent with read lock on read open", + /* 8 */"Read lease gets SIGIO on write open", }; static int64_t tests[][5] = @@ -195,6 +197,26 @@ static int64_t tests[][5] = {6, CMD_CLOSE, 0, PASS, SERVER }, {6, CMD_CLOSE, 0, PASS, CLIENT }, + /* Don't get SIGIO when read lease is taken by read */ + {7, CMD_OPEN, O_RDONLY, PASS, CLIENT }, + {7, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, + {7, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, + {7, CMD_SIGIO, 0, PASS, CLIENT }, + {7, CMD_OPEN, O_RDONLY, PASS, SERVER }, + {7, CMD_WAIT_SIGIO, 5, FAIL, CLIENT }, + {7, CMD_CLOSE, 0, PASS, SERVER }, + {7, CMD_CLOSE, 0, PASS, CLIENT }, + + /* Get SIGIO when Read lease is broken by Write */ + {8, CMD_OPEN, O_RDONLY, PASS, CLIENT }, + {8, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, + {8, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, + {8, CMD_SIGIO, 0, PASS, CLIENT }, + {8, CMD_OPEN, O_RDWR, PASS, SERVER }, + {8, CMD_WAIT_SIGIO, 5, PASS, CLIENT }, + {8, CMD_CLOSE, 0, PASS, SERVER }, + {8, CMD_CLOSE, 0, PASS, CLIENT }, + /* indicate end of array */ /* Must have an end for the SERVER and one for the CLIENT */ {0,0,0,0,SERVER},
On Tue, 2019-09-03 at 15:22 -0700, Ira Weiny wrote: > On Wed, Sep 04, 2019 at 07:42:55AM +1000, Dave Chinner wrote: > > On Tue, Sep 03, 2019 at 02:05:35PM -0700, ira.weiny@intel.com wrote: > > > From: Ira Weiny <ira.weiny@intel.com> > > > > > > To get to a point of developing and testing the new Lease semantics for long > > > term file pins[1], add a simple lease test. > > > > Hi Ira, > > > > Can you co-ordinate with Jeff on this? He just posted a patch > > to add F_SETLEASE testing with a new binary a few days > > ago... > > > > https://marc.info/?l=fstests&m=156690934513424&w=2 > > Sure. > > Jeff, > > Looking at your submission revealed I was missing 2 tests (shown in the diff > below). > > Would it be sufficient to add these to this submission and go with my series? > Or would you prefer using fork vs the socket harness? > > I think we may need a fork in the server side of my tests to work on duplicated > FD's but right now I have more tests than you. > > Let me know, Thanks, > Ira > > commit 849633871b7c0b00a6ee2bb751c09cd6365da386 (lease-test-b0-v3) > Author: Ira Weiny <ira.weiny@intel.com> > Date: Tue Sep 3 15:17:52 2019 -0700 > > src/leasetest: Add tests to match competing test > > A competing patch to test leases was submitted in parallel.[1] > > This adds tests which were covered in that patch set. > > [1] https://marc.info/?l=fstests&m=156690934513424&w=2 > > Signed-off-by: Ira Weiny <ira.weiny@intel.com> > > diff --git a/src/leasetest.c b/src/leasetest.c > index b30a4d05d039..729aca851493 100644 > --- a/src/leasetest.c > +++ b/src/leasetest.c > @@ -137,6 +137,8 @@ char *descriptions[] = { > /* 4 */"Fail Read Lease if opened with write permissions", > /* 5 */"Read lease gets SIGIO on write open", > /* 6 */"Write lease gets SIGIO on read open", > + /* 7 */"No SIGIO is sent with read lock on read open", > + /* 8 */"Read lease gets SIGIO on write open", > }; > > static int64_t tests[][5] = > @@ -195,6 +197,26 @@ static int64_t tests[][5] = > {6, CMD_CLOSE, 0, PASS, SERVER }, > {6, CMD_CLOSE, 0, PASS, CLIENT }, > > + /* Don't get SIGIO when read lease is taken by read */ > + {7, CMD_OPEN, O_RDONLY, PASS, CLIENT }, > + {7, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, > + {7, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, > + {7, CMD_SIGIO, 0, PASS, CLIENT }, > + {7, CMD_OPEN, O_RDONLY, PASS, SERVER }, > + {7, CMD_WAIT_SIGIO, 5, FAIL, CLIENT }, > + {7, CMD_CLOSE, 0, PASS, SERVER }, > + {7, CMD_CLOSE, 0, PASS, CLIENT }, > + > + /* Get SIGIO when Read lease is broken by Write */ > + {8, CMD_OPEN, O_RDONLY, PASS, CLIENT }, > + {8, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, > + {8, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, > + {8, CMD_SIGIO, 0, PASS, CLIENT }, > + {8, CMD_OPEN, O_RDWR, PASS, SERVER }, > + {8, CMD_WAIT_SIGIO, 5, PASS, CLIENT }, > + {8, CMD_CLOSE, 0, PASS, SERVER }, > + {8, CMD_CLOSE, 0, PASS, CLIENT }, > + > /* indicate end of array */ > /* Must have an end for the SERVER and one for the CLIENT */ > {0,0,0,0,SERVER}, Hi Ira, Let's go with your testcase, including the patch above. It's more comprehensive than the one I proposed, and should be easier to extend in the future. Acked-by: Jeff Layton <jlayton@kernel.org>
> Subject: Re: [RFC PATCH 0/2] Add lease testing > > On Tue, 2019-09-03 at 15:22 -0700, Ira Weiny wrote: > > On Wed, Sep 04, 2019 at 07:42:55AM +1000, Dave Chinner wrote: > > > On Tue, Sep 03, 2019 at 02:05:35PM -0700, ira.weiny@intel.com wrote: > > > > From: Ira Weiny <ira.weiny@intel.com> > > > > > > > > To get to a point of developing and testing the new Lease > > > > semantics for long term file pins[1], add a simple lease test. > > > > > > Hi Ira, > > > > > > Can you co-ordinate with Jeff on this? He just posted a patch to add > > > F_SETLEASE testing with a new binary a few days ago... > > > > > > https://marc.info/?l=fstests&m=156690934513424&w=2 > > > > Sure. > > > > Jeff, > > > > Looking at your submission revealed I was missing 2 tests (shown in > > the diff below). > > > > Would it be sufficient to add these to this submission and go with my > series? > > Or would you prefer using fork vs the socket harness? > > > > I think we may need a fork in the server side of my tests to work on > > duplicated FD's but right now I have more tests than you. > > > > Let me know, Thanks, > > Ira > > > > commit 849633871b7c0b00a6ee2bb751c09cd6365da386 (lease-test-b0-v3) > > Author: Ira Weiny <ira.weiny@intel.com> > > Date: Tue Sep 3 15:17:52 2019 -0700 > > > > src/leasetest: Add tests to match competing test > > > > A competing patch to test leases was submitted in parallel.[1] > > > > This adds tests which were covered in that patch set. > > > > [1] https://marc.info/?l=fstests&m=156690934513424&w=2 > > > > Signed-off-by: Ira Weiny <ira.weiny@intel.com> > > > > diff --git a/src/leasetest.c b/src/leasetest.c index > > b30a4d05d039..729aca851493 100644 > > --- a/src/leasetest.c > > +++ b/src/leasetest.c > > @@ -137,6 +137,8 @@ char *descriptions[] = { > > /* 4 */"Fail Read Lease if opened with write permissions", > > /* 5 */"Read lease gets SIGIO on write open", > > /* 6 */"Write lease gets SIGIO on read open", > > + /* 7 */"No SIGIO is sent with read lock on read open", > > + /* 8 */"Read lease gets SIGIO on write open", > > }; > > > > static int64_t tests[][5] = > > @@ -195,6 +197,26 @@ static int64_t tests[][5] = > > {6, CMD_CLOSE, 0, PASS, SERVER }, > > {6, CMD_CLOSE, 0, PASS, CLIENT }, > > > > + /* Don't get SIGIO when read lease is taken by read */ > > + {7, CMD_OPEN, O_RDONLY, PASS, CLIENT }, > > + {7, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, > > + {7, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, > > + {7, CMD_SIGIO, 0, PASS, CLIENT }, > > + {7, CMD_OPEN, O_RDONLY, PASS, SERVER }, > > + {7, CMD_WAIT_SIGIO, 5, FAIL, CLIENT }, > > + {7, CMD_CLOSE, 0, PASS, SERVER }, > > + {7, CMD_CLOSE, 0, PASS, CLIENT }, > > + > > + /* Get SIGIO when Read lease is broken by Write */ > > + {8, CMD_OPEN, O_RDONLY, PASS, CLIENT }, > > + {8, CMD_SETLEASE, F_RDLCK, PASS, CLIENT }, > > + {8, CMD_GETLEASE, F_RDLCK, PASS, CLIENT }, > > + {8, CMD_SIGIO, 0, PASS, CLIENT }, > > + {8, CMD_OPEN, O_RDWR, PASS, SERVER }, > > + {8, CMD_WAIT_SIGIO, 5, PASS, CLIENT }, > > + {8, CMD_CLOSE, 0, PASS, SERVER }, > > + {8, CMD_CLOSE, 0, PASS, CLIENT }, > > + > > /* indicate end of array */ > > /* Must have an end for the SERVER and one for the CLIENT */ > > {0,0,0,0,SERVER}, > > Hi Ira, > > Let's go with your testcase, including the patch above. It's more > comprehensive than the one I proposed, and should be easier to extend in > the future. > > Acked-by: Jeff Layton <jlayton@kernel.org> Thanks! Give me a bit and I will submit V2 with this squashed into the original patch. I did some cleanup of locktest but other than that leasetest is pretty much a copy. Not sure if the community would like to see more changes or not. Ira
From: Ira Weiny <ira.weiny@intel.com> To get to a point of developing and testing the new Lease semantics for long term file pins[1], add a simple lease test. Currently the following tests are performed based on the fcntl() man page. /* 1 */"Take Read Lease", /* 2 */"Take Write Lease", /* 3 */"Fail Write Lease if file is open somewhere else", /* 4 */"Fail Read Lease if opened with write permissions", /* 5 */"Read lease gets SIGIO on write open", /* 6 */"Write lease gets SIGIO on read open", [1] https://lkml.org/lkml/2019/8/14/435 Ira Weiny (2): src/leasetest: Add lease test executable generic: Add Lease testing src/Makefile | 2 +- src/leasetest.c | 837 ++++++++++++++++++++++++++++++++++++++++++ tests/generic/567 | 100 +++++ tests/generic/567.out | 2 + tests/generic/group | 1 + 5 files changed, 941 insertions(+), 1 deletion(-) create mode 100644 src/leasetest.c create mode 100755 tests/generic/567 create mode 100644 tests/generic/567.out