Page MenuHomeFreeBSD

kevans (Kyle Evans)
User

Projects (9)

User Details

User Since
Oct 2 2015, 1:17 PM (441 w, 3 d)

Recent Activity

Yesterday

kevans committed rG97bfe58ddc60: crunchgen: slap a dependency on the generated makefile for .lo (authored by kevans).
crunchgen: slap a dependency on the generated makefile for .lo
Mon, Mar 18, 3:53 PM
kevans committed rGefb74e16ddc0: libsysdecode: add a decoder for pollfd events (authored by kevans).
libsysdecode: add a decoder for pollfd events
Mon, Mar 18, 3:53 PM
kevans committed rG160788caa3c9: ktrace: log genio events on failed write (authored by kevans).
ktrace: log genio events on failed write
Mon, Mar 18, 3:53 PM
kevans committed rGb983d4f63cb5: kdump: decode pollfd struct arrays coming from poll(2) (authored by kevans).
kdump: decode pollfd struct arrays coming from poll(2)
Mon, Mar 18, 3:53 PM
kevans committed rG416547ba79c3: kern: poll: tap out the pollfd array on successful return (authored by kevans).
kern: poll: tap out the pollfd array on successful return
Mon, Mar 18, 3:53 PM
kevans committed rGa7b90458a293: rtld: add some dlopen tests (authored by kevans).
rtld: add some dlopen tests
Mon, Mar 18, 3:53 PM
kevans added a comment to D44401: daemon(8): handle case of waitpid() returning without exited child.

Oh, return () style differs here

Mon, Mar 18, 2:42 PM
kevans accepted D44401: daemon(8): handle case of waitpid() returning without exited child.
Mon, Mar 18, 2:41 PM
kevans accepted D44407: libbe: Fix some markup issues..
Mon, Mar 18, 1:36 PM
kevans accepted D44406: bectl: Simplify command aliases..
Mon, Mar 18, 1:35 PM
kevans accepted D44405: bectl: Use geopt() and drop mention of -?..
Mon, Mar 18, 1:35 PM
kevans committed rG79a095fa8a66: crunchgen: slap a dependency on the generated makefile for .lo (authored by kevans).
crunchgen: slap a dependency on the generated makefile for .lo
Mon, Mar 18, 3:42 AM
kevans committed rG7e9323311117: rtld: add some dlopen tests (authored by kevans).
rtld: add some dlopen tests
Mon, Mar 18, 3:42 AM

Fri, Mar 15

kevans committed rG3705d679a634: if_wg: use proper barriers around pkt->p_state (authored by kevans).
if_wg: use proper barriers around pkt->p_state
Fri, Mar 15, 1:26 AM
kevans closed D44283: if_wg: use proper barriers around pkt->p_state.
Fri, Mar 15, 1:26 AM

Tue, Mar 12

kevans added a comment to D44200: tun(4)/tap(4): allow devices to be configured as transient.

So, with D44307 in, you can add this:

static struct if_clone *
tuntap_cloner_from_flags(int tun_flags)
{
 
        for (u_int i = 0; i < NDRV; i++)
                if ((tun_flags & TUN_DRIVER_IDENT_MASK) ==
                    tuntap_drivers[i].ident_flags)
                        return (V_tuntap_driver_cloners[i]);
 
        return (NULL);
}

And then call

if_clone_destroyif(tuntap_cloner_from_flags(flags), ifp);

This of course won't cover the problematic traveling case. To cover that, you'd need to:

CURVNET_SET_QUIET(ifp->if_home_vnet);
cloner = tuntap_cloner_from_flags(flags)
CURVNET_RESTORE();
if_clone_destroyif(cloner, ifp);
Tue, Mar 12, 5:36 AM
kevans accepted D44307: if_tuntap: simplify storage of per-vnet cloners.

Nice, thanks. I think at one point I thought I might have something else worth adding to struct tuntap_driver_cloner, but if it hasn't happened in the intervening 5 years then I think it's safe to say that it's not going to happen.

Tue, Mar 12, 5:31 AM
kevans added inline comments to D44283: if_wg: use proper barriers around pkt->p_state.
Tue, Mar 12, 5:04 AM

Mon, Mar 11

kevans added a comment to D44200: tun(4)/tap(4): allow devices to be configured as transient.

Damn, this again comes to the stupid idea of interfaces traveling between vnets. Grrrr. Still, I believe it should be done in the tuntap, not in if_clone.c. The existing SLIST of cloners can be simplified a lot, let me try.

Mon, Mar 11, 6:27 PM
kevans added inline comments to D44200: tun(4)/tap(4): allow devices to be configured as transient.
Mon, Mar 11, 5:53 PM
kevans added inline comments to D44283: if_wg: use proper barriers around pkt->p_state.
Mon, Mar 11, 2:56 PM

Sat, Mar 9

kevans closed D43869: crunchgen: slap a dependency on the generated makefile for .lo.
Sat, Mar 9, 2:02 AM
kevans committed rG6e2cfb24ee87: crunchgen: slap a dependency on the generated makefile for .lo (authored by kevans).
crunchgen: slap a dependency on the generated makefile for .lo
Sat, Mar 9, 2:02 AM
kevans added inline comments to D44283: if_wg: use proper barriers around pkt->p_state.
Sat, Mar 9, 1:53 AM
kevans requested review of D44283: if_wg: use proper barriers around pkt->p_state.
Sat, Mar 9, 1:34 AM

Thu, Mar 7

kevans accepted D44183: libc/iconv: don't export nonexistant symbols.

ah, I see; some of these were read from a header, some of them left un-exported in the first place... someone seemingly applied a relatively large hammer.

Thu, Mar 7, 2:39 AM

Tue, Mar 5

kevans added inline comments to D44167: script: Handle terminal resize on SIGSWINCH.
Tue, Mar 5, 2:39 PM
kevans updated the diff for D43921: pci: avoid accidental clobbering of regs on some fdt platforms.

Revert back to the size check, assert that we do or do not have a resource
after as appropriate.

Tue, Mar 5, 5:51 AM
kevans closed D43799: [RFC] ktrace: log genio events on failed write.
Tue, Mar 5, 5:46 AM
kevans committed rG47ad4f2d45e4: ktrace: log genio events on failed write (authored by kevans).
ktrace: log genio events on failed write
Tue, Mar 5, 5:46 AM
kevans closed D44160: kdump: decode pollfd struct arrays coming from poll(2).
Tue, Mar 5, 5:46 AM
kevans committed rG02c57f7b4877: kdump: decode pollfd struct arrays coming from poll(2) (authored by kevans).
kdump: decode pollfd struct arrays coming from poll(2)
Tue, Mar 5, 5:46 AM
kevans closed D44159: libsysdecode: add a decoder for pollfd events.
Tue, Mar 5, 5:46 AM
kevans committed rGbd23e71f91ce: libsysdecode: add a decoder for pollfd events (authored by kevans).
libsysdecode: add a decoder for pollfd events
Tue, Mar 5, 5:46 AM
kevans closed D44158: kern: poll: tap out the pollfd array on successful return.
Tue, Mar 5, 5:45 AM
kevans committed rGb5d2165b9046: kern: poll: tap out the pollfd array on successful return (authored by kevans).
kern: poll: tap out the pollfd array on successful return
Tue, Mar 5, 5:45 AM
kevans added a comment to D44160: kdump: decode pollfd struct arrays coming from poll(2).

Oh, that probably won't work... we probably printed the entire left-hand side before the return, so it'd need to be on a separate line or something

Tue, Mar 5, 4:26 AM
kevans added a comment to D44160: kdump: decode pollfd struct arrays coming from poll(2).
In D44160#1008533, @jhb wrote:

Next you might consider adding this to truss as well. :)

Tue, Mar 5, 4:24 AM

Mon, Mar 4

kevans added inline comments to D44167: script: Handle terminal resize on SIGSWINCH.
Mon, Mar 4, 4:06 PM

Sun, Mar 3

kevans requested review of D44199: kern: net: allow an if_clone_destroyif() without knowing the cloner.
Sun, Mar 3, 8:20 PM
kevans requested review of D44200: tun(4)/tap(4): allow devices to be configured as transient.
Sun, Mar 3, 8:20 PM
kevans updated the diff for D39740: if_tuntap: make SIOCIFDESTROY interruptible.

Use a loop, set error before (it'll either get clobbered or remain untouched),
and just consistently keep it locked leaving the loop.

Sun, Mar 3, 7:16 PM
kevans added a comment to D39740: if_tuntap: make SIOCIFDESTROY interruptible.

(Sorry, I just realized I forgot about this review entirely)

Sun, Mar 3, 7:15 PM
kevans accepted D44195: Add POSIX_SPAWN_DISABLE_ASLR_NP.

Second, the setting is inherited by children of the spawn process. Is it fine? For me yes.

Sun, Mar 3, 2:45 PM

Sat, Mar 2

kevans added inline comments to D44183: libc/iconv: don't export nonexistant symbols.
Sat, Mar 2, 1:21 AM

Fri, Mar 1

kevans added inline comments to D44167: script: Handle terminal resize on SIGSWINCH.
Fri, Mar 1, 4:05 PM

Thu, Feb 29

kevans added inline comments to D44159: libsysdecode: add a decoder for pollfd events.
Thu, Feb 29, 4:50 PM
kevans requested review of D44160: kdump: decode pollfd struct arrays coming from poll(2).
Thu, Feb 29, 4:47 PM
kevans requested review of D44159: libsysdecode: add a decoder for pollfd events.
Thu, Feb 29, 4:47 PM
kevans requested review of D44158: kern: poll: tap out the pollfd array on successful return.
Thu, Feb 29, 4:47 PM

Wed, Feb 28

kevans added a comment to D44141: nuageinit: add basic support for cloudinit..

I'd recommend doing a linting pass with devel/lua-luacheck- usually the concerns it points out aren't too crazy.

Wed, Feb 28, 3:57 PM
kevans accepted D44143: loader/lua: Remove compat shim for loader.lua_path.
Wed, Feb 28, 3:17 PM
kevans accepted D44142: loader/lua: Remove pager shim.
Wed, Feb 28, 3:16 PM
kevans accepted D44144: loader/lua: Remove workaround for command_error.
Wed, Feb 28, 3:16 PM

Mon, Feb 26

kevans accepted D44094: loader: Add loader.exit.
Mon, Feb 26, 11:54 PM
kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.
In D44030#1005630, @dim wrote:

@devnexen_gmail.com I think you implemented https://github.com/llvm/llvm-project/commit/691b12a2dcc12fa43517d23f2a9b6039616eebc8, but the problem is that it specifcally did _not_ call elf_aux_info since that is intercepted, and might lead to segfaults.

Mon, Feb 26, 2:04 AM

Sun, Feb 25

kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.
In D44030#1005624, @dim wrote:

Have you tried running the lsan test suite with this change?

I'm working on it, but just trying to get the llvm test suite running has been... rough. I chose to redirect from your question to the tangentially related and still valuable: "Have you tried running the asan test suite with this change?" since we can know how much of the asan test suite has passed already.

After fumbling around and finally figuring out that I needed to configure llvm/ with -DLLVM_ENABLE_PROJECTS="clang;compiler-rt" because some of the test prereqs are in llvm or clang, most of the tests that actually run fail because they're trying to do something like:

22:56 <@kevans91_>  70252 103012 Asan-aarch64-inline CALL  execve(0,0x81278f78,0x81278f98)
22:56 <@kevans91_>  70252 103012 Asan-aarch64-inline RET   execve -1 errno 9 Bad file descriptor

... and this is the realization that I just burned a not-insignificant chunk of time because this hasn't been upstreamed yet, three months later: https://cgit.freebsd.org/src/commit/?id=4c9a0adad18263ec8725d9bfc5f560c6ad1da8bd

If you or @dim upload that change I can give it an approval and we can merge it.

It was committed past november, in https://github.com/llvm/llvm-project/commit/7440e4ed85aa9 , which was a fold-up of:

Sun, Feb 25, 8:21 PM

Sat, Feb 24

kevans accepted D44062: loader/efi: Use unique linker set for lua.
Sat, Feb 24, 11:40 PM
kevans accepted D44061: loader/efi: Linker sets not needed..

s/matater/matter/ in commit msg

Sat, Feb 24, 11:40 PM
kevans accepted D44060: loader/kboot: simplify linker set inclusion a little.
Sat, Feb 24, 11:39 PM
kevans accepted D44059: loader/uboot: Move to foo.ldconfig.
Sat, Feb 24, 11:38 PM
kevans accepted D44058: loader/powerpc: Share ldscript.
Sat, Feb 24, 11:38 PM
kevans accepted D44057: kboot: Centralize ldscript addition.
Sat, Feb 24, 11:38 PM
kevans accepted D44056: loader: Move ldscripts to match more standard practices.
Sat, Feb 24, 11:38 PM
kevans accepted D44063: loader/ficl: Rename the ficl compile set to X4th_compile_set.
Sat, Feb 24, 11:37 PM
kevans accepted D44064: loader/efi: Only include interpreter's linker script.
Sat, Feb 24, 11:37 PM
kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.

The lsan tests reliably trigger the deadlock described above LockStuffAndStopTheWorld, so I can't really get any valid results from that one.

Sat, Feb 24, 7:05 AM
kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.

Have you tried running the lsan test suite with this change?

Sat, Feb 24, 5:57 AM

Fri, Feb 23

kevans added a comment to D44041: lldb lua: Add some helper functions to deal with netinet.

lldb multiple scripting interfaces, most notably in python and lua

Fri, Feb 23, 4:59 PM
kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.

My hope was to get a sanity check on StopTheWorld, at least, beforehand since that's a FreeBSD implementation detail and it seems easier to collect reviews for that specifically here.

Fri, Feb 23, 2:51 PM
kevans added a comment to D44030: llvm: implement the bits missing for LSan on FreeBSD.

Oh, and I note that I've only tested this on amd64 and arm64 to make sure that both the rfork_thread and rfork paths work. LSan is incredibly useful when used in an ASan + libFuzzer combination.

Fri, Feb 23, 2:12 AM
kevans requested review of D44031: lib: libclang_rt: hook up lsan where appropriate.
Fri, Feb 23, 2:04 AM
kevans requested review of D44030: llvm: implement the bits missing for LSan on FreeBSD.
Fri, Feb 23, 2:04 AM

Thu, Feb 22

kevans accepted D44016: kboot: Implement write support for hostdisk.

Ah, yeah, that'd definitely do it...

Thu, Feb 22, 12:29 AM

Feb 17 2024

kevans added a comment to D43921: pci: avoid accidental clobbering of regs on some fdt platforms.
In D43921#1002445, @jhb wrote:

This is probably fine, but I think you could also just leave the rid member off. I'm working on a series that removes the rid argument from bus_release_resource. My one thought for having the rid in the ranges structure is if we wanted to set the rids and do bus_set_resource in the bus-specific attach routine when the ranges are enumerated, but I'm not sure it's worth moving the code there (where it would have to be duplicated).

Feb 17 2024, 2:50 AM
kevans updated the diff for D43921: pci: avoid accidental clobbering of regs on some fdt platforms.

Stash the rid off in the pcie_range, use that everywhere after allocation both
for bus_*_resource() APIs bus also to decide if we're skipping an element to
deallocate.

Feb 17 2024, 12:34 AM

Feb 15 2024

kevans updated the diff for D43921: pci: avoid accidental clobbering of regs on some fdt platforms.

Add a macro, fixup other places, commentary

Feb 15 2024, 6:36 PM
kevans requested review of D43921: pci: avoid accidental clobbering of regs on some fdt platforms.
Feb 15 2024, 5:42 PM
kevans accepted D43908: loader: Move drawer.lua over to gfx table..
Feb 15 2024, 12:43 AM
kevans accepted D43907: loader: Move gfx functions to gfx.lua.8.
Feb 15 2024, 12:42 AM
kevans accepted D43906: loader: Move to using linker sets to bring in optional bits.
Feb 15 2024, 12:41 AM
kevans accepted D43905: loader: Remove gfx_fb_stub.c, it's no longer needed.
Feb 15 2024, 12:41 AM
kevans accepted D43904: loader: Only create gfx 4th bindings when gfx is available.
Feb 15 2024, 12:40 AM
kevans accepted D43903: loader: Add prototype for gfx_interp_md.
Feb 15 2024, 12:40 AM
kevans accepted D43902: loader: Create new gfx table.

Hah

Feb 15 2024, 12:39 AM
kevans accepted D43901: loader: Separate gfx to a new file..
Feb 15 2024, 12:38 AM
kevans accepted D43900: loader: Register the gfx stuff separately..
Feb 15 2024, 12:38 AM

Feb 14 2024

kevans added a comment to D43858: More avoidance of the filtees loading recursion.

hm, I applied this one https://people.freebsd.org/~krion/D43858.diff and it worked

This is definitively different from what I have... but it is also different from what I get displayed here in the review.

Feb 14 2024, 2:40 PM
kevans added inline comments to D43872: flua: Add hash module.
Feb 14 2024, 3:43 AM

Feb 13 2024

kevans committed rG16eebc4e19de: caroot: routine update (authored by kevans).
caroot: routine update
Feb 13 2024, 8:35 PM
kevans committed rG9fc1d78e39d5: caroot: routine update (authored by kevans).
caroot: routine update
Feb 13 2024, 7:17 PM
kevans committed rG9b7611d9c7b4: caroot: routine update (authored by kevans).
caroot: routine update
Feb 13 2024, 7:16 PM
kevans requested review of D43869: crunchgen: slap a dependency on the generated makefile for .lo.
Feb 13 2024, 7:04 PM
kevans committed rGc5796f1572c8: rtld: add some dlopen tests (authored by kevans).
rtld: add some dlopen tests
Feb 13 2024, 3:39 PM
kevans closed D43859: rtld: add some dlopen tests.
Feb 13 2024, 3:39 PM
kevans added a comment to D43859: rtld: add some dlopen tests.
In D43859#1000568, @kib wrote:

How fresh are processes where ATF executes tests? Would both dlopen_basic and dlopen_recursing execute in the same image sequentially, or ATF spawns/execs a new process to run each test?

Feb 13 2024, 3:55 AM
kevans added a comment to D43858: More avoidance of the filtees loading recursion.

Relevant test in https://reviews.freebsd.org/D43859

Feb 13 2024, 2:31 AM
kevans requested review of D43859: rtld: add some dlopen tests.
Feb 13 2024, 2:31 AM

Feb 12 2024

kevans accepted D43810: libsecureboot do not report expected unverified files.

No comment on the veriexec bits (they seem generally sane, but that's your area of course :-)), but the motivation described here:

Feb 12 2024, 9:49 PM
kevans accepted D43844: rescue: belatedly add zfsbootcfg.
Feb 12 2024, 5:15 PM