- User Since
- Oct 16 2018, 11:01 AM (145 w, 1 d)
Apr 28 2021
Apr 26 2021
Apr 17 2021
Added a test to show that we don't report POLLRDHUP unless you asked for it.
Ok, now ignoring POLLINIGNEOF for this.
Apr 14 2021
This time with -U999999, and mentioning illumos as requested.
Jan 11 2021
Jan 9 2021
Jan 8 2021
Looks OK to me. I suppose we could also use bitmask style for LIO_SYNC and LIO_DATA_ONLY?
Jan 6 2021
Updated to address review feedback and fix oversights in man page.
Rebased. Previous confusion turned out to be explained by problems fixed by D27353 and problems in my test methodology.
Rebased (mostly: ZFS moved).
Dec 16 2020
Great feature. +1 for allowing it in lio_listio() for batched submissions in later work (I'd also like to see LIO_FSYNC allowed in there too). I wonder if it is OK to use C11 anon union syntax in a system header; I see that sys/ucred.h does that.
Dec 10 2020
Dec 9 2020
Dropped the SIGEV_THREAD function pointer. Switched to a hardcoded limit on how many lio_listio array elements to display, following a coding pattern from nearby similar things.
Dec 8 2020
Nov 25 2020
Nov 24 2020
Hmm. My filesystem has 32kb blocks, and this generates reads, according to dwatch -X io:
Nov 8 2020
Jul 12 2020
Jun 23 2020
Jun 21 2020
Jun 17 2020
Jun 2 2020
Jun 1 2020
Handled feedback from asomers@: added a test, added a missing check in aio_queue_file(), used -U999999 for more context. I also created a separate review ticket D25090 to add O_DSYNC for open(2) and fcntl(2). The present patch is rebased on top of that one.
May 30 2020
This seems correct at a high level. It's required to read the data back. (For what it's worth, failure to distinguish between data integrity meta-data relating to file extension and frippery like mtime seems to have been a problem on other popular OSs' fdatasync(2) implementations in the past, which I think is what leads Postgres to trust fdatasync(2) only when it knows there has been a full fsync(2) since the last file size change. It would be cool to be able to save on some I/Os be being less paranoid.)
May 27 2020
Dec 13 2019
Discussed with Mateusz, he would prefer to write a Coccinelle script to automate this change, so I am abandoning this patch.
Oct 4 2019
Aug 19 2019
Thanks for the reminder. I will commit this later today.
Aug 2 2019
Ok, based on feedback here and elsewhere, here's a version that uses getopt_long() from libc. It supports both --ppid (like Linux ps) and -P (like pgrep), and it seemed reasonable to add obvious long opts for --pid, --tty and --jail too. I decided not to try to add Linux-like --user, --User, --group, --Group in this patch (-u, -U, -g, -G all look a bit unportable and complicated). Thoughts?
Jul 30 2019
Jul 18 2019
Tested, the bad code removal works as described. +1 for getting rid of the BACKWARD zones by default, to match base. LGTM.
Jul 17 2019
Jul 15 2019
Jun 29 2019
I suppose we could update tzsetup(8) to use the new version too, if we're going to install it. So here's a patch that teaches that to handle the modern format with multi-country first field, and gets rid of the hardcoded "Arctic" menu because it went away in the data (concretely that means that people in https://en.wikipedia.org/wiki/Svalbard would need to choose Europe -> Norway, whose timezone they share as I assume they are well aware, being "a full part of the Kingdom of Norway").
Jun 15 2019
Jun 6 2019
Please re-read what I suggested in previous reply.
May 25 2019
Got rid of allocation/copying.
Show POSIX segments as "(unlinked)" if their name has gone.
Show SysV segments with their key_t (same way it is displayed in ipcs output).
May 23 2019
An obvious example of another user would be SysV shm. I suppose it should show the key (in hex format?) as the "path". Then of course you would use jail_adjust = false.
May 22 2019
May 20 2019
May 18 2019
Thanks for working on this! I took it for a quick spin and have some initial feedback.
Feb 28 2019
Feb 22 2019
Nov 20 2018
Oct 28 2018
Oct 18 2018
Oct 16 2018
Here's a better version (the last one had an unintended stray change).
Note: this is a refactoring suggested by mjg, to support his future plans to change the representation of panic state to something small in __read_frequently. The new macro will hide those details from all these call sites, though for now it just does what it was doing before (with the addition of the prediction).