Pending 3.20 release

Pending 3.20 release

[Jens Axboe]

Hi,

CC'ing a few folks here that have had (or have) pending patches
that aren't yet in the repo. I'm planning on releasing 3.20 in
the not-so-distant future, but I wanted to check in and see if
everybody is happy with where we are at for their favorite features
or bug fixes.

Let me know if you have pending items that should go into 3.20,
or if you're happy with the current state.

Thanks,
-- 
Jens Axboe

Re: Pending 3.20 release

[Alexey Dobriyan]

On Thu, May 21, 2020 at 08:40:09AM -0600, Jens Axboe wrote:
> Hi,
> 
> CC'ing a few folks here that have had (or have) pending patches
> that aren't yet in the repo. I'm planning on releasing 3.20 in
> the not-so-distant future, but I wanted to check in and see if
> everybody is happy with where we are at for their favorite features
> or bug fixes.
> 
> Let me know if you have pending items that should go into 3.20,
> or if you're happy with the current state.

I'll rebase and resend everything except ZBD verification.

bugfix			fio: parse "io_size=1%"
			zbd: consolidate zone mutex initialisation
			zbd: make zbd_info->mutex non-recursive
quite important feature	zbd: introduce per job maximum open zones limit
more assert()		zbd: don't lock zones outside working area
			zbd: bump ZBD_MAX_OPEN_ZONES
			verify: decouple seed generation from buffer fill

Re: Pending 3.20 release

[Jens Axboe]

On 5/21/20 11:28 AM, Alexey Dobriyan wrote:
> On Thu, May 21, 2020 at 08:40:09AM -0600, Jens Axboe wrote:
>> Hi,
>>
>> CC'ing a few folks here that have had (or have) pending patches
>> that aren't yet in the repo. I'm planning on releasing 3.20 in
>> the not-so-distant future, but I wanted to check in and see if
>> everybody is happy with where we are at for their favorite features
>> or bug fixes.
>>
>> Let me know if you have pending items that should go into 3.20,
>> or if you're happy with the current state.
> 
> I'll rebase and resend everything except ZBD verification.
> 
> bugfix			fio: parse "io_size=1%"
> 			zbd: consolidate zone mutex initialisation
> 			zbd: make zbd_info->mutex non-recursive
> quite important feature	zbd: introduce per job maximum open zones limit
> more assert()		zbd: don't lock zones outside working area
> 			zbd: bump ZBD_MAX_OPEN_ZONES
> 			verify: decouple seed generation from buffer fill

Great thanks, your series is the only one I'm aware of that is pending,
so was hoping you'd pipe up.

-- 
Jens Axboe

Re: Pending 3.20 release

[Rebecca Cran]

[-- Attachment #1: Type: text/plain, Size: 3688 bytes --]

On 5/21/20 8:40 AM, Jens Axboe wrote:

> Let me know if you have pending items that should go into 3.20,
> or if you're happy with the current state.

I've not run fio on FreeBSD for a while, and just noticed a couple of 
things - which fortunately aren't blockers since I see you've already 
tagged 3.20 :)

Running "./fio --help" prints the usage information and then pauses for 
about a second. Interrupting it and looking at the backtrace shows:


(gdb) bt
#0  _umtx_op () at _umtx_op.S:4
#1  0x00000008007260b0 in pshared_gc (curthread=0x801212000) at 
/usr/src/lib/libthr/thread/thr_pshared.c:110
#2  __thr_pshared_destroy (key=<optimized out>) at 
/usr/src/lib/libthr/thread/thr_pshared.c:253
#3  0x0000000800720023 in _thr_mutex_destroy (mutex=0x801809bd8) at 
/usr/src/lib/libthr/thread/thr_mutex.c:482
#4  0x0000000000268d97 in __fio_sem_remove (sem=0x801809bd8) at fio_sem.c:20
#5  0x000000000028db35 in fio_filelock_exit () at filelock.c:104
#6  0x0000000000244ead in free_shm () at init.c:335
#7  0x000000080091be25 in __cxa_finalize (dso=0x0) at 
/usr/src/lib/libc/stdlib/atexit.c:240
#8  0x00000008008aea71 in exit (status=0) at 
/usr/src/lib/libc/stdlib/exit.c:74
#9  0x000000000024495c in parse_cmd_line (argc=2, argv=0x7fffffffeaa0, 
client_type=1) at init.c:2901
#10 0x0000000000244f02 in parse_options (argc=2, argv=0x7fffffffeaa0) at 
init.c:2958
#11 0x000000000029828a in main (argc=2, argv=0x7fffffffeaa0, 
envp=<optimized out>) at fio.c:42


Also, building on FreeBSD or Linux with clang 10.0 produces lots of 
warnings. I've uploaded the build log to 
http://bex.dev/typescript-fio320-clang10-linux .

An example of the warnings:

crc/../arch/../lib/rand.h:109:25: warning: implicit conversion from 
'unsigned long long' to 'double' changes value from 18446744073709551615 
to 18446744073709551616 [-Wimplicit-int-float-conversion]
                 return (val + 1.0) / (FRAND64_MAX + 1.0);
                                       ^~~~~~~~~~~ ~
crc/../arch/../lib/rand.h:9:22: note: expanded from macro 'FRAND64_MAX'
#define FRAND64_MAX     (-1ULL)
                          ^~~~~
crc/../arch/../lib/rand.h:136:41: warning: implicit conversion from 
'unsigned long long' to 'double' changes value from 18446744073709551615 
to 18446744073709551616 [-Wimplicit-int-float-conversion]
         return (uint64_t) ((double)end * (r / (FRAND64_MAX + 1.0)));
                                                ^~~~~~~~~~~ ~
crc/../arch/../lib/rand.h:9:22: note: expanded from macro 'FRAND64_MAX'
#define FRAND64_MAX     (-1ULL)
                          ^~~~~
2 warnings generated.
crc/murmur3.c:33:2: warning: unannotated fall-through between switch 
labels [-Wimplicit-fallthrough]
         case 2:
         ^
crc/murmur3.c:33:2: note: insert '__attribute__((fallthrough));' to 
silence this warning
         case 2:
         ^
         __attribute__((fallthrough));
crc/murmur3.c:33:2: note: insert 'break;' to avoid fall-through
         case 2:
         ^
         break;
crc/murmur3.c:36:2: warning: unannotated fall-through between switch 
labels [-Wimplicit-fallthrough]
         case 1:
         ^
crc/murmur3.c:36:2: note: insert '__attribute__((fallthrough));' to 
silence this warning
         case 1:
         ^
         __attribute__((fallthrough));
crc/murmur3.c:36:2: note: insert 'break;' to avoid fall-through
         case 1:
         ^
         break;

-- 
Rebecca Cran


[-- Attachment #2: typescript-clang10-freebsd13 --]
[-- Type: application/octet-stream, Size: 191031 bytes --]

Re: Pending 3.20 release

[Tomohiro Kusumi]

2020年5月26日(火) 13:03 Rebecca Cran <rebecca@bsdio.com>:
>
> On 5/21/20 8:40 AM, Jens Axboe wrote:
>
> > Let me know if you have pending items that should go into 3.20,
> > or if you're happy with the current state.
>
> I've not run fio on FreeBSD for a while, and just noticed a couple of
> things - which fortunately aren't blockers since I see you've already
> tagged 3.20 :)
>
> Running "./fio --help" prints the usage information and then pauses for
> about a second. Interrupting it and looking at the backtrace shows:

I've git-bisected this.
The delay was caused by this commit which first appeared in fio-3.17.
Note that fio on NetBSD didn't show this delay.

commit 43661a66a8c3b9fb5ff28b463c9ab0aac91a0355 (refs/bisect/bad)
Date:   Tue Oct 8 16:13:02 2019 -0700
    Increase MAX_FILELOCKS for highly parallel IO test


>
>
> (gdb) bt
> #0  _umtx_op () at _umtx_op.S:4
> #1  0x00000008007260b0 in pshared_gc (curthread=0x801212000) at
> /usr/src/lib/libthr/thread/thr_pshared.c:110
> #2  __thr_pshared_destroy (key=<optimized out>) at
> /usr/src/lib/libthr/thread/thr_pshared.c:253
> #3  0x0000000800720023 in _thr_mutex_destroy (mutex=0x801809bd8) at
> /usr/src/lib/libthr/thread/thr_mutex.c:482
> #4  0x0000000000268d97 in __fio_sem_remove (sem=0x801809bd8) at fio_sem.c:20
> #5  0x000000000028db35 in fio_filelock_exit () at filelock.c:104
> #6  0x0000000000244ead in free_shm () at init.c:335
> #7  0x000000080091be25 in __cxa_finalize (dso=0x0) at
> /usr/src/lib/libc/stdlib/atexit.c:240
> #8  0x00000008008aea71 in exit (status=0) at
> /usr/src/lib/libc/stdlib/exit.c:74
> #9  0x000000000024495c in parse_cmd_line (argc=2, argv=0x7fffffffeaa0,
> client_type=1) at init.c:2901
> #10 0x0000000000244f02 in parse_options (argc=2, argv=0x7fffffffeaa0) at
> init.c:2958
> #11 0x000000000029828a in main (argc=2, argv=0x7fffffffeaa0,
> envp=<optimized out>) at fio.c:42
>
>
> Also, building on FreeBSD or Linux with clang 10.0 produces lots of
> warnings. I've uploaded the build log to
> http://bex.dev/typescript-fio320-clang10-linux .
>
> An example of the warnings:
>
> crc/../arch/../lib/rand.h:109:25: warning: implicit conversion from
> 'unsigned long long' to 'double' changes value from 18446744073709551615
> to 18446744073709551616 [-Wimplicit-int-float-conversion]
>                  return (val + 1.0) / (FRAND64_MAX + 1.0);
>                                        ^~~~~~~~~~~ ~
> crc/../arch/../lib/rand.h:9:22: note: expanded from macro 'FRAND64_MAX'
> #define FRAND64_MAX     (-1ULL)
>                           ^~~~~
> crc/../arch/../lib/rand.h:136:41: warning: implicit conversion from
> 'unsigned long long' to 'double' changes value from 18446744073709551615
> to 18446744073709551616 [-Wimplicit-int-float-conversion]
>          return (uint64_t) ((double)end * (r / (FRAND64_MAX + 1.0)));
>                                                 ^~~~~~~~~~~ ~
> crc/../arch/../lib/rand.h:9:22: note: expanded from macro 'FRAND64_MAX'
> #define FRAND64_MAX     (-1ULL)
>                           ^~~~~
> 2 warnings generated.
> crc/murmur3.c:33:2: warning: unannotated fall-through between switch
> labels [-Wimplicit-fallthrough]
>          case 2:
>          ^
> crc/murmur3.c:33:2: note: insert '__attribute__((fallthrough));' to
> silence this warning
>          case 2:
>          ^
>          __attribute__((fallthrough));
> crc/murmur3.c:33:2: note: insert 'break;' to avoid fall-through
>          case 2:
>          ^
>          break;
> crc/murmur3.c:36:2: warning: unannotated fall-through between switch
> labels [-Wimplicit-fallthrough]
>          case 1:
>          ^
> crc/murmur3.c:36:2: note: insert '__attribute__((fallthrough));' to
> silence this warning
>          case 1:
>          ^
>          __attribute__((fallthrough));
> crc/murmur3.c:36:2: note: insert 'break;' to avoid fall-through
>          case 1:
>          ^
>          break;
>
> --
> Rebecca Cran
>