User Details
- User Since
- Jul 22 2016, 2:56 AM (539 w, 2 d)
Yesterday
Sat, Nov 14
- Removed FreeBSD-specific comment; motivation for defining a max need not be prescribed
In the previous change where I added this check, I was trying to avoid adding a FreeBSD ifdef in the generic code. But after adding the NM_GENERIC_MAX_BUF_SIZE definition, that works nicely as a platform-agnostic opt-in mechanism.
MFC 3/3 in stack of errata candidates.
This is 2/3 in a stack of errata candidates for stable/15.
Note: I've adjusted the Fixes: label from the original commit message to refer to the stable/15 commit here.
This is the first of three commits in the stack. These are errata candidates.
Thu, Nov 12
I didn't look at quite enough context and missed that the variables are already declared above the loop and used further down below it. I created a new diff, since git-arc didn't want to update this one.
D61455
Wed, Nov 11
- Dropped unneeded __restrict
Mon, Nov 9
Addressed feedback:
- Use a more specific macro name for the max mbuf size for emulation
I've used uint32_t to match the type of the ext_size and objsize fields the values come from.
Incorporated feedback:
- Make size unsigned in nm_os_mbuf_reinit()
Incorporated feedback:
- Make nm_os_get_mbuf() take a uint32_t size, matching the type of the field passed to it.
- Define and use NETMAP_MAX_BUF_SIZE
Sat, Nov 7
Wed, Nov 4
Small nit: the code after looks to be easily simplified by using strlcpy() in place of strlen() + bcopy() while here.
The added null check is in very deep and we should probably not be getting so far into VBE code with no VBE support. This seems it should instead be guarded using vbe_check() early in vbe_default_mode() as is done in other public functions here.
A question this brought up while looking into it: the lib/liblua/luaconf.h file seems stale and unused, and I wonder if it is similar for stand/liblua/luaconf.h? The build seems to be using contrib/lua/src/luaconf.h, which is the copy that actually includes this lib/liblua/luaconf.local.h file.
As an example of where this can be useful, it enables io.popen() to invoke a command with a bidirectional pipe by passing "r+" or "r+e" as the mode. My use case is for wrapping "cxgbetool stdio" to
Oct 13 2026
Oct 10 2026
With this change I can run zdb (and other libzpool-based tooling) compiled for Linux under Linuxulator.
The added commit is related, though Phabricator does not make that clear.
Oct 9 2026
Oct 7 2026
libfetch has fetchLastErrString and fetchLastErrCode you can use to return more specific error information than errno provides.
Feb 8 2023
Feb 7 2023
Feb 2 2023
Sep 12 2022
If snapshots_changed property has local source for a filesystem or volume, this kernel patch will remove it after a snapshot is created or destroyed on that dataset. To list the affected datasets:
Aug 30 2022
Aug 20 2022
Aug 17 2022
Aug 16 2022
Jul 19 2022
Jun 27 2022
Would you mind opening a PR at https://github.com/openzfs/zfs for this?
Jun 9 2022
Jun 1 2022
May 25 2022
Thanks for the patch. The fse and zstd warnings have been fixed in OpenZFS master already and I have just opened a pull request to fix the dnode warning. That way we can avoid patching. I will update the port when that goes in.
Apr 4 2022
Mar 9 2022
Feb 14 2022
Jan 28 2022
Jan 26 2022
Jan 21 2022
Jan 20 2022
Dec 30 2021
Dec 15 2021
Nov 16 2021
@rmacklem is it ok if we make parsecred return an int so we can fail to start when the creds are invalid? It feels wrong to merely log an error message and keep going when this is misconfigured.