- User Since
- Nov 24 2013, 3:15 AM (222 w, 1 d)
Other than a couple of small comments I think this is good to go.
MI - Machine Independent
MD - Machine Dependent (i.e., architecture-specific)
Fri, Feb 23
I'd suggest we commit this incrementally; I'm looking at the MD/MI movement atm.
@trasz sorry I accidentally committed r329877 but referenced this review; can you please re-upload?
Committed as r329877 (which accidentally reported review D3715).
Thu, Feb 22
I think this looks good. (Note that I haven't yet reviewed in detail.)
Wed, Feb 21
- rename KERNEL_RETPOLINE to separate it from possible future userland support
- default to no for now, with the intent of setting it in GENERIC as is done with CTF
Tue, Feb 20
Sat, Feb 17
Fri, Feb 16
After r329370 remaining files waiting on approval from:
- Assar Westerlund
- Doug Rabson
- Peter Wemm
- Tim J. Robbins
- Dmitry Chagin
Refresh after partial commit
Thu, Feb 15
Here's the acpidump -dt:
Tue, Feb 13
Yes. At least the plan of record is we ditch GCC4.2 in 13. That will be nice.
I tested net/samba47 builds as suggested in PR 225584. Previous tests without this D14347 patch (but some with various diagnostic or workaround patches) hung between the first and 20th or so iteration. With this patch applied my Threadripper completed 132 cycles of net/samba47 build last night.
Mon, Feb 12
Sun, Feb 11
Sat, Feb 10
Testing with this change on X260 (LEN0068, MHKV 0x0200 AFAICT) I see warnings like:
# sysctl dev.acpi_ibm.0.kbd_backlight=0 ACPI Warning: \_SB.PCI0.LPC.EC.HKEY.MLCG: Insufficient arguments - Caller passed 0, method requires 1 (20180209/nsarguments-383) ACPI Warning: \_SB.PCI0.LPC.EC.HKEY.MLCG: Insufficient arguments - Caller passed 0, method requires 1 (20180209/nsarguments-383) dev.acpi_ibm.0.kbd_backlight: 1ACPI Warning: \_SB.PCI0.LPC.EC.HKEY.MLCG: Insufficient arguments - Caller passed 0, method requires 1 (20180209/nsarguments-383) -> ACPI Warning: \_SB.PCI0.LPC.EC.HKEY.MLCG: Insufficient arguments - Caller passed 0, method requires 1 (20180209/nsarguments-383) ACPI Warning: \_SB.PCI0.LPC.EC.HKEY.MLCG: Insufficient arguments - Caller passed 0, method requires 1 (20180209/nsarguments-383) 0 #
Keyboard backlight control works despite the warnings. I haven't tested this machine on an unpatched kernel. echo 1 > /dev/led/micmute has no effect.
I presume the commit will have Sponsored bys for Google Summer of Code, Netflix, and the FreeBSD Foundation; any others?
No regressions observed on my Thinkpad X220 with this change.
Changing dev.acpi_ibm.0.volume has no effect, but did not work before either.
I'll test this on old and hopefully new thinkpads
Fri, Feb 9
Microoptimize away a fork+exec of tr, from @stevek
Thu, Feb 8
This works in my QEMU test case and LGTM, modulo indentation nits.
Existing indentation in this file is a mess, but there are a couple of new cases that don't match either of the two existing styles. I suggest tweaking this change to match the surrounding cases, and then someone (@imp or me, perhaps) should follow up with reformatting the portions that don't conform to style(9).
Wed, Feb 7
With this patch I see six register-indirect jmps/calls in a GENERIC kernel:
ffffffff80999047: ff d0 callq *%rax ffffffff80e3c0d6: ff d0 callq *%rax ffffffff80e5d5d0: ff 14 24 callq *(%rsp) ffffffff80f3d79f: ff 55 f8 callq *-8(%rbp) ffffffff80f88b40: ff 54 24 10 callq *16(%rsp) ffffffff80f88b60: ff 54 24 10 callq *16(%rsp)
The corresponding symbols for these are: privcmd_ioctl, outofnmi, sigcode, hypercall_md, ia32_sigcode, freebsd4_ia32_sigcode.
Use COMPILER_FEATURES not LINKER_FEATURES for testing compiler features as reported by @dim.
Example of retpoline, from sys/cam/scsi/scsi_enc.c
These changes will be included in the work to upstream the man page, tracked in https://reviews.llvm.org/D42963
LGTM, would like to see ideally @bdrewery but at least someone else agree
Tue, Feb 6
LGTM although I still think a comment is warranted
Quote COMPILER_FEATURES for CROSSENV as it may contain spaces
Might be wroth adding a comment that find -s is not portable, to discourage someone from changing it back to remove the extra exec as an optimization.
improvements from kib