cem (Conrad Meyer)
User

Projects

User Details

User Since
Jul 9 2015, 9:56 PM (154 w, 2 d)

Recent Activity

Yesterday

cem committed rS335584: aesni(4): Fix {de,en}crypt operations that allocated a buffer.
aesni(4): Fix {de,en}crypt operations that allocated a buffer
Sat, Jun 23, 6:20 PM
cem committed rS335583: aesni(4): Support CRD_F_KEY_EXPLICIT OCF mode.
aesni(4): Support CRD_F_KEY_EXPLICIT OCF mode
Sat, Jun 23, 5:24 PM
cem committed rP473112: Update devel/elfutils to 0.172.
Update devel/elfutils to 0.172
Sat, Jun 23, 3:56 AM

Fri, Jun 22

cem added a comment to D15560: Use a spin lock to serialize updates on AMD CPUs..

Apologies for the delay. Would anyone with a Ryzen or TR be able to test a microcode update with the updated patch applied?

Fri, Jun 22, 4:39 PM

Thu, Jun 21

cem added a comment to D15952: Fix the Linux kernel version number calculation.

Yeah, I buy that argument. We should check for non-negative but otherwise be permissive.

Thu, Jun 21, 8:41 PM
cem added inline comments to D15952: Fix the Linux kernel version number calculation.
Thu, Jun 21, 8:13 PM
cem accepted D15952: Fix the Linux kernel version number calculation.

LGTM modulo one nit below

Thu, Jun 21, 8:04 PM
cem added a comment to D15952: Fix the Linux kernel version number calculation.

I don't know where it lives in the Linux source tree, but it ends up installed as /usr/include/linux/version.h with a numerically identical definition to the existing one in compat/linux.

Thu, Jun 21, 6:38 PM
cem committed rS335498: Sync strlcpy with userland version, again.
Sync strlcpy with userland version, again
Thu, Jun 21, 5:35 PM
cem added a comment to D15952: Fix the Linux kernel version number calculation.

I don't think there's any real reason to diverge from how Linux computes its own version numbers, just to be "more native" FreeBSD. I don't know if these numbers are exposed in any userspace interface, but if they are obviously they need to match Linux, not our own made-up scheme.

Thu, Jun 21, 4:44 PM
cem requested changes to D15952: Fix the Linux kernel version number calculation.

Why not leave the macro alone and compute v using it instead?

Thu, Jun 21, 4:23 PM

Wed, Jun 20

cem added a comment to D15905: safer wait-free iteration of shared interrupt handlers.
In D15905#337247, @avg wrote:

Additionally, anything "interrupt thread" is completely irrelevant to interrupt filters that can also share an interrupt.

Wed, Jun 20, 9:14 PM
cem added a comment to D15905: safer wait-free iteration of shared interrupt handlers.
In D15905#337246, @avg wrote:
In D15905#337244, @cem wrote:

Maybe we could allocate independent ithreads for each handler on a shared interrupt source. I'm not sure what the ramifications of that are or if there is a good reason we do not already do that. It may just be that shared interrupts mostly date to the UP days.

This was implemented in the code that you recently removed (INTR_FILTER).

Wed, Jun 20, 9:13 PM
cem added a comment to D15905: safer wait-free iteration of shared interrupt handlers.
In D15905#337213, @avg wrote:

I am not sure what exactly you mean by this.
What kind of contention?

Wed, Jun 20, 9:06 PM
cem added a comment to D15905: safer wait-free iteration of shared interrupt handlers.
In D15905#337055, @avg wrote:
In D15905#337019, @cem wrote:

CK_SLIST when most of accesses are done under a lock is an unneeded overhead.

Can you elaborate on that overhead?

I mean that even if the list is accessed under a mutex the CK_SLIST operations would still use all the fences because they wouldn't know about the mutex.
Or if the code already has fences around a larger block that includes CL_LIST operations.
Maybe it's a trivial cost and nothing to worry about.

Wed, Jun 20, 5:40 PM
cem added a comment to D15905: safer wait-free iteration of shared interrupt handlers.

CK_SLIST when most of accesses are done under a lock is an unneeded overhead.

Wed, Jun 20, 4:13 PM

Tue, Jun 19

cem added a comment to D15560: Use a spin lock to serialize updates on AMD CPUs..

If anyone with a Ryzen that has SMT enabled could print logical_cpus_mask (e.g., from kgdb) and paste it here, I'd appreciate it. I'd prefer to use that instead of a spin lock. Otherwise I will just go ahead and commit this in the next day or two, since it's simple and easily MFCed.

Tue, Jun 19, 9:24 PM

Mon, Jun 18

cem resigned from D15886: Expose ZFS quotas over nfs.
Mon, Jun 18, 8:10 PM

Fri, Jun 15

cem committed rD51838: Document __FreeBSD_version 1200069.
Document __FreeBSD_version 1200069
Fri, Jun 15, 9:08 PM
cem committed rS335237: Bump __FreeBSD_version after r335227.
Bump __FreeBSD_version after r335227
Fri, Jun 15, 8:57 PM
cem committed rS335227: Retain offset compatibility with pre-12.0 dumps.
Retain offset compatibility with pre-12.0 dumps
Fri, Jun 15, 7:03 PM
cem closed D15829: Retain offset compatibility with pre-12.0 dumps.
Fri, Jun 15, 7:03 PM
cem added a comment to D15829: Retain offset compatibility with pre-12.0 dumps.

I think this is fine. The panic string buffer is bigger than it needs to be.

We already have the compile-time assert in all the MD dump files.

Fri, Jun 15, 6:57 PM
cem added a dependency for D15829: Retain offset compatibility with pre-12.0 dumps: D11723: Add support for compressed kernel dumps..
Fri, Jun 15, 6:45 PM
cem created D15829: Retain offset compatibility with pre-12.0 dumps.
Fri, Jun 15, 6:45 PM
Herald added a reviewer for D11723: Add support for compressed kernel dumps.: manpages.
Fri, Jun 15, 6:38 PM

Sun, Jun 10

cem added inline comments to D15711: uuencode(1): Capsicumify.
Sun, Jun 10, 12:34 AM

Wed, Jun 6

cem committed rS334738: strdup.3: Document standardization in POSIX.
strdup.3: Document standardization in POSIX
Wed, Jun 6, 9:33 PM
cem committed rS334733: strcpy.3: Improve legibility and clarity.
strcpy.3: Improve legibility and clarity
Wed, Jun 6, 8:09 PM
cem committed rS334722: puc(4): Add provisional support for Exar XR17V352.
puc(4): Add provisional support for Exar XR17V352
Wed, Jun 6, 4:47 PM
cem committed rS334721: clog.3, complex.3: Fix typos and igor style issues.
clog.3, complex.3: Fix typos and igor style issues
Wed, Jun 6, 4:25 PM

Mon, Jun 4

cem committed rS334625: Correctly handle the padding for IPv6-AH, as specified by RFC4302.
Correctly handle the padding for IPv6-AH, as specified by RFC4302
Mon, Jun 4, 6:51 PM
cem committed rS334624: str(r)chr: Replace union abuse with __DECONST.
str(r)chr: Replace union abuse with __DECONST
Mon, Jun 4, 6:47 PM
cem created D15661: Correctly handle the padding for IPv6-AH, as specified by RFC4302.
Mon, Jun 4, 4:08 AM

Sat, Jun 2

cem added a comment to D15526: reduce overhead of entropy collection.
In D15526#330724, @jmg wrote:

Oh, note that the above are my opinions, and they differ significantly from most of the rest of the FreeBSD team. Mainly why I haven't spent much time working on the rng system,

Sat, Jun 2, 8:54 PM
cem created P184 mjg #s 2018-06-02.
Sat, Jun 2, 6:55 PM
cem added a comment to D15526: reduce overhead of entropy collection.
In D15526#330700, @jmg wrote:

as I have suggested, doing a sampling is a good choice for those that want/need it.. i.e. use 10 packets/second (or less). Use the previous second's pps to calculate the n for selecting every nth packet to use such that it comes out to something like 10pps.

Sat, Jun 2, 6:20 PM

Fri, Jun 1

cem committed rS334503: at.man: Bump .Dd missed in r334502.
at.man: Bump .Dd missed in r334502
Fri, Jun 1, 10:57 PM
cem committed rS334502: Update other man pages to match leap second reality.
Update other man pages to match leap second reality
Fri, Jun 1, 10:38 PM
cem committed rS334501: touch.1: Update to conform to POSIX 2004.
touch.1: Update to conform to POSIX 2004
Fri, Jun 1, 10:35 PM

Thu, May 31

cem added a comment to D15607: Improve error handling for KERN_PROC_{CWD,FILEDESC}..

Exactly, and the error should be propagated up so that the right layer (e.g., sysctl vs. core dump) can decide whether to throw it away or not. As it happens, the core dump code already ignores these errors, so this change is for the benefit of sysctl interfaces.

Thu, May 31, 11:40 PM
cem added a comment to D15607: Improve error handling for KERN_PROC_{CWD,FILEDESC}..
In D15607#330347, @jhb wrote:

I do think it was intentional to truncate once the output buffer was exhausted rather than failing with ENOMEM. Its debatable if that is a useful thing or not. Most sysctls return ENOMEM instead so that the request can be retried with a larger output buffer so it is probably better to do what this does.

Thu, May 31, 9:49 PM
cem committed rS334443: dhclient(8): allow to supersede interface-mtu option.
dhclient(8): allow to supersede interface-mtu option
Thu, May 31, 7:36 PM
cem closed D15484: dhclient: allow to supersede interface-mtu option.
Thu, May 31, 7:36 PM
cem accepted D15623: sysutils/fusefs-lkl: update to 4.16.0.
Thu, May 31, 3:44 PM
cem accepted D15526: reduce overhead of entropy collection.
Thu, May 31, 3:42 AM

Wed, May 30

cem accepted D15623: sysutils/fusefs-lkl: update to 4.16.0.
Wed, May 30, 11:09 PM
cem added a comment to D15526: reduce overhead of entropy collection.
In D15526#330095, @jeff wrote:

Given that there is trivially little if any entropy coming from mbufs is there a reason we're leaving this callsite at all? has anyone from secteam commented?

Wed, May 30, 11:09 PM
cem accepted D15526: reduce overhead of entropy collection.

LGTM modulo nit about only disabling on amd64.

Wed, May 30, 10:37 PM
cem accepted D15623: sysutils/fusefs-lkl: update to 4.16.0.

Looks fine to me, other than the one nit below. You will need a port committer's approval to get the change committed.

Wed, May 30, 8:04 PM
cem requested changes to D10778: [concurrency_kit] turn this into a module!.

It's used in core kernel code now. To limit size damage, it may be possible to restrict what is built-in on smaller platforms (e.g., do we need anything other than epoch support code?). But I think a module is no longer a useful option, since it will need to be loaded everywhere.

Wed, May 30, 5:30 PM

Tue, May 29

cem accepted D15607: Improve error handling for KERN_PROC_{CWD,FILEDESC}..

Looks fine to me, with the caveat that I haven't looked at what might break with an error return. I'm most interested in the impact on core dumps.

Tue, May 29, 8:39 PM
cem added inline comments to D15607: Improve error handling for KERN_PROC_{CWD,FILEDESC}..
Tue, May 29, 8:05 PM
cem added a comment to D15607: Improve error handling for KERN_PROC_{CWD,FILEDESC}..

I don't see how this affects tmux in any way, unless the error code breaks the previously working-but-truncated path(s). If that's the case, I think those should be changed to swallow ENOMEM. But then, what's the point of propagating the error up?

Tue, May 29, 7:58 PM
cem added inline comments to D15574: cpucontrol: fix debugging for family on AMD cpus and add useful debugging..
Tue, May 29, 1:44 AM

May 25 2018

cem accepted D15574: cpucontrol: fix debugging for family on AMD cpus and add useful debugging..

LGTM modulo my first comment.

May 25 2018, 5:26 PM
cem added a comment to D15523: sysutils/devcpu-data: Update AMD microcode..

Hrm... Can you run it with my diff for more debugging goodness? https://people.freebsd.org/~sbruno/cpucontrol.diff.txt

May 25 2018, 5:19 PM
cem added a comment to D15523: sysutils/devcpu-data: Update AMD microcode..

cpucontrol print on multiple input files could be a lot more clear :(. I disambiguated it by copying fam17h to its own directory:

May 25 2018, 4:41 PM
cem added a comment to D15567: amdtemp and amdsmn update for new devices.

No worries. I'll continue tinker with this. Do you have a 0x1463 device (in the same machine is 0x1450)?

May 25 2018, 4:28 PM
cem added a comment to D15567: amdtemp and amdsmn update for new devices.

I don't have any doc for the new Zen APUs, sorry. My TR doesn't have any 15eb devices.

May 25 2018, 3:39 PM
cem added a comment to D15389: aesni(4): improve session lookup performance.

So what is the plan here? Wait for somebody willing to do the session management in OCF right?

May 25 2018, 7:21 AM · secteam

May 24 2018

cem added a comment to D15560: Use a spin lock to serialize updates on AMD CPUs..
In D15560#328553, @avg wrote:

I changed the update code to what it is now (synchronized update), because on my Phenom II based systems the old update mode, core after core, driven by userland, caused hard hangs.
My assumption at the time was that the microcode can affect the inter-core communication (or HyperTransport), so it was necessary to keep all core safely "parked" and to update them at the same time.
I didn't have access to any non-public AMD documentation, so I don't know what really happened and what we were supposed to do, but that change fixed my problem and didn't seem to introduce any new problems.

May 24 2018, 8:41 PM
cem accepted D15560: Use a spin lock to serialize updates on AMD CPUs..

Patch seems harmless.

May 24 2018, 8:36 PM
cem added a comment to D15560: Use a spin lock to serialize updates on AMD CPUs..

Here's what I've gotten out of the Epyc PPR:

May 24 2018, 8:28 PM
cem committed rS334170: Yank crufty INTR_FILTER option.
Yank crufty INTR_FILTER option
May 24 2018, 5:06 PM
cem closed D15546: Yank crufty INTR_FILTER option.
May 24 2018, 5:06 PM
cem added a comment to D15546: Yank crufty INTR_FILTER option.

(rpaulo was the original author of the driver (and the INTR_FILTER cruft dates to the original commit), but he is inactive now.)

May 24 2018, 4:35 PM
cem added a comment to D15389: aesni(4): improve session lookup performance.

Nobody wants this? It is definitely better than the current code though.

May 24 2018, 3:03 PM · secteam
cem created D15546: Yank crufty INTR_FILTER option.
May 24 2018, 3:49 AM

May 23 2018

cem added inline comments to D15526: reduce overhead of entropy collection.
May 23 2018, 12:54 AM

May 19 2018

cem accepted D15484: dhclient: allow to supersede interface-mtu option.

@cem I'd appreciate if you could get this committed for me.

May 19 2018, 7:00 AM
cem accepted D15484: dhclient: allow to supersede interface-mtu option.
May 19 2018, 6:11 AM

May 18 2018

cem accepted D15484: dhclient: allow to supersede interface-mtu option.
May 18 2018, 8:58 PM

May 17 2018

cem added a comment to D15465: Enable kernel dump options on most architectures..

It looks like the GENERIC configs that you're not updating are for ARM, RISC-V, and PPC64. Why are they being excluded?

May 17 2018, 11:52 PM
cem accepted D15465: Enable kernel dump options on most architectures..
May 17 2018, 5:58 PM

May 16 2018

cem committed rS333683: teken: Unbreak syscons' use of teken.
teken: Unbreak syscons' use of teken
May 16 2018, 6:13 PM

May 14 2018

cem removed a reviewer for D15429: Use /lib:/usr/lib as default libs in mips64.: cem.

Seems harmless but I have no interest in MIPS.

May 14 2018, 4:32 PM

May 11 2018

cem added a reviewer for D15389: aesni(4): improve session lookup performance: jhb.

I agree that the current system sucks and that it does not scale with multiple sessions. I'm not sure a binary tree is the right replacement, though. And probably session management should be lifted out into OCF (OpenCrypto Framework) so that all drivers can benefit from it without copy-pasting.

May 11 2018, 4:09 PM · secteam

May 4 2018

cem accepted D15300: iflib: print message when iflib_tx_structures_setup fails.
May 4 2018, 6:57 PM

Apr 28 2018

cem committed rS333087: amd64/mp_machdep.c: Fix GCC build after r333059.
amd64/mp_machdep.c: Fix GCC build after r333059
Apr 28 2018, 5:55 PM
cem committed rS333086: smartpqi(4): Use conventional path to driver sources.
smartpqi(4): Use conventional path to driver sources
Apr 28 2018, 5:44 PM
cem committed rS333085: smartpqi(4): Unbreak GCC build.
smartpqi(4): Unbreak GCC build
Apr 28 2018, 5:43 PM

Apr 27 2018

cem accepted D15203: Post-panic KASSERT enhancements.

Looks fine to me.

Apr 27 2018, 12:51 AM

Apr 24 2018

cem committed rS332962: Do not totally silence suppressed secondary kasserts unless debug.kassert..
Do not totally silence suppressed secondary kasserts unless debug.kassert.
Apr 24 2018, 7:11 PM
cem committed rS332960: Fix debug.kassert.do_log description text.
Fix debug.kassert.do_log description text
Apr 24 2018, 6:59 PM
cem committed rS332959: panic: Optionally, trace secondary panics.
panic: Optionally, trace secondary panics
Apr 24 2018, 6:54 PM
cem committed rS332957: lockmgr: Add missed neutering during panic.
lockmgr: Add missed neutering during panic
Apr 24 2018, 6:41 PM

Apr 21 2018

cem committed rS332853: blake2: Disable warnings (not just error) for code we will not modify.
blake2: Disable warnings (not just error) for code we will not modify
Apr 21 2018, 2:09 AM

Apr 19 2018

cem committed rS332778: cryptosoft: Do not exceed crd_len around *crypt_multi.
cryptosoft: Do not exceed crd_len around *crypt_multi
Apr 19 2018, 3:24 PM

Apr 17 2018

cem accepted D15119: Handle overflow of uid or gid in arguments for chown.

Next time, please upload a diff with more context. Either use diff -U999999 to generate patches, or the arc command to directly upload changesets to Phab.

Apr 17 2018, 8:53 PM · Contributor Reviews (base)
cem added inline comments to D15115: top: fix warnings from clang/gcc.
Apr 17 2018, 5:48 PM
cem committed rS332655: libmd: Remove trailing whitespace from mdXhl.c.
libmd: Remove trailing whitespace from mdXhl.c
Apr 17 2018, 5:23 PM

Apr 16 2018

cem committed rS332632: quota(1): Fix calculation overflow and underflow.
quota(1): Fix calculation overflow and underflow
Apr 16 2018, 7:33 PM

Apr 14 2018

cem committed rS332499: expr(1): Fix overflow detection when operand is INTMAX_MIN.
expr(1): Fix overflow detection when operand is INTMAX_MIN
Apr 14 2018, 4:35 AM
cem committed rS332498: cpufreq: Remove error-prone table terminators in favor of automatic sizing.
cpufreq: Remove error-prone table terminators in favor of automatic sizing
Apr 14 2018, 3:15 AM

Apr 10 2018

cem accepted D15018: pf: Limit ioctl to a reasonable and tuneable number of elements.
In D15018#316088, @cem wrote:

Looks good in general. It seems that some cases use WOULD_OVERFLOW as a secondary check and some don't. Is there any reason for that?

Good catch, I missed it in the first two cases because that used to be a fixed 65k value. Now that it's user configurable it needs the check too.

Apr 10 2018, 4:04 PM

Apr 9 2018

cem accepted D15000: Allow to verify keys in geli..
Apr 9 2018, 8:41 PM

Apr 8 2018

cem added inline comments to D15010: add white listing for ZFS locking pairs that WITNESS can't report accurately and enable WITNESS by default in ZFS.
Apr 8 2018, 11:04 PM
cem accepted D15018: pf: Limit ioctl to a reasonable and tuneable number of elements.

Looks good in general. It seems that some cases use WOULD_OVERFLOW as a secondary check and some don't. Is there any reason for that?

Apr 8 2018, 9:55 PM
cem added inline comments to D15000: Allow to verify keys in geli..
Apr 8 2018, 4:56 PM