IABSD.fr/src/sys
Branch :
Log
| Author | Commit | Date | CI | Message |
|---|---|---|---|---|
 |
4b0a39dc | 2025-07-12 00:41:14 | refcnt_shared can be a macro around refcnt_read. they're basically the same code except refcnt_shared checks if the read value was > 1. the caller is going to check it anyway, so get rid of the duplicate code. | |
 |
b3bc833a | 2025-07-11 20:04:20 | AMD SEV-ES guest: In vctrap() allow only CPUID from userspace. CPUID is the only instruction we allow to raise a #VC exception also from user space. All other instructions are limited to raise #VC from kernel space only. With respect to rdmsr/wrmsr, this is an additional safe-guard, as these two instructions will raise a #GP anyway when the CPL is greater than 0. With respect to in/out, userland could be allowed to access IO ports. However, our SEV-ES guest do not support this for now. from hshoexer@; OK mlarkin@ | |
 |
36cc6724 | 2025-07-11 19:12:49 | Remove unused uuid_dec_{be,le} routines. Last uses were removed with subr_disk.c r1.275. | |
 |
7674e898 | 2025-07-11 01:17:42 | avoid unused variable warning on gcc4 | |
|
8ba5287a | 2025-07-11 00:55:08 | drm/amdkfd: Fix instruction hazard in gfx12 trap handler From Jay Cornwall 725a59d29adb517cc197763edbe91606fc564d6f in linux-6.12.y/6.12.36 424648c3838133f93a34fdfe4f9d5597551e7b3b in mainline linux | |
|
55ed7455 | 2025-07-11 00:53:27 | drm/amdkfd: remove gfx 12 trap handler page size cap From Jonathan Kim 6c7dc7ad867c14638cadfe9c0a2e93d2507002c7 in linux-6.12.y/6.12.36 cd82f29ec51b2e616289db7b258a936127c16efa in mainline linux | |
|
41ecc0f3 | 2025-07-11 00:51:33 | drm/fbdev-dma: Add shadow buffering for deferred I/O From Thomas Zimmermann 0d087de947babf7ed70029d042abcc6ed06ff415 in linux-6.12.y/6.12.36 3603996432997f7c88da37a97062a46cda01ac9d in mainline linux | |
|
068991ae | 2025-07-11 00:49:27 | drm/amd/display: Fix mpv playback corruption on weston From Alex Hung 8cd7ee9cd7decf195a4fa0098a1a0308d85f9ad9 in linux-6.12.y/6.12.36 8724a5380c4390eed81e271d22f34ff06453ded9 in mainline linux | |
|
7a6e2c31 | 2025-07-11 00:47:02 | drm/amdgpu: switch job hw_fence to amdgpu_fence From Alex Deucher 5f2e040f19c4d73496ac8c32f2c441a6146430c9 in linux-6.12.y/6.12.36 ebe43542702c3d15d1a1d95e8e13b1b54076f05a in mainline linux | |
|
5199a02a | 2025-07-11 00:43:54 | drm/amdgpu: Fix SDMA UTC_L1 handling during start/stop sequences From Jesse Zhang 9cfa2fea2566fc488aa59e80ea00422c3e441e47 in linux-6.12.y/6.12.36 7f3b16f3f229e37cc3e02e9e4e7106c523b119e9 in mainline linux | |
|
d036958b | 2025-07-11 00:41:56 | drm/i915/dsi: Fix off by one in BXT_MIPI_TRANS_VTOTAL From Ville Syrjala cc0a3fd781bf83d631d988568871808d06046e75 in linux-6.12.y/6.12.36 c464ce6af332e7c802c36cd337cacf81db05400c in mainline linux | |
|
acae1e28 | 2025-07-11 00:40:02 | drm/amd/display: Check dce_hwseq before dereferencing it From Alex Hung e881b82f5d3d8d54d168cd276169f0fee01bf0e7 in linux-6.12.y/6.12.36 b669507b637eb6b1aaecf347f193efccc65d756e in mainline linux | |
|
6eb0554e | 2025-07-11 00:38:18 | drm/amdgpu: Add kicker device detection From Frank Min 593517e5561cb8f7f03503de3d75169b855b1d00 in linux-6.12.y/6.12.36 0bbf5fd86c585d437b75003f11365b324360a5d6 in mainline linux | |
|
52e73eb7 | 2025-07-11 00:35:50 | drm/amd/display: Fix RMCM programming seq errors From Yihan Zhu ba1ffc32bda798613999916c60a03462233ae067 in linux-6.12.y/6.12.36 158f9944ac05dafd2d3a23d0688e6cf40ef68b90 in mainline linux | |
|
2b8682ba | 2025-07-11 00:33:49 | drm/amd/display: Correct non-OLED pre_T11_delay. From Zhongwei Zhang ca8efc6a89716245408eb00a6a83dfa13dea7a2f in linux-6.12.y/6.12.36 893f07452bca56ff146a6be02b3294a9ea23d18a in mainline linux | |
|
68ceec27 | 2025-07-11 00:30:48 | drm/amdgpu: amdgpu_vram_mgr_new(): Clamp lpfn to total vram From John Olender e2c3133ff4d594f68d3bae82b2637e96b509e063 in linux-6.12.y/6.12.36 4d2f6b4e4c7ed32e7fa39fcea37344a9eab99094 in mainline linux | |
|
f447154d | 2025-07-11 00:29:06 | drm/amd/display: Add null pointer check for get_first_active_display() From Wentao Liang 4ce9f2dc9ff7cc410e8c5d936ec551e26b9599a9 in linux-6.12.y/6.12.36 c3e9826a22027a21d998d3e64882fa377b613006 in mainline linux | |
|
184c5cc3 | 2025-07-11 00:27:18 | drm/amdkfd: Fix race in GWS queue scheduling From Jay Cornwall b681e2a8a7595431479a3743ef2e148f79cbe4b5 in linux-6.12.y/6.12.36 cfb05257ae168a0496c7637e1d9e3ab8a25cbffe in mainline linux | |
|
9d63b3f6 | 2025-07-11 00:25:40 | drm/i915: fix build error some more From Arnd Bergmann 18ec560e3e9b476e63c671514dcb8b9163a578a2 in linux-6.12.y/6.12.36 d02b2103a08b6d6908f1d3d8e8783d3f342555ac in mainline linux | |
|
e542b685 | 2025-07-11 00:24:02 | drm/amd: Adjust output for discovery error handling From Mario Limonciello 23116bf9a3d046fb0057cb6a121f97219dce4607 in linux-6.12.y/6.12.36 73eab78721f7b85216f1ca8c7b732f13213b5b32 in mainline linux | |
|
a69ebe02 | 2025-07-11 00:22:14 | drm/amdgpu/discovery: optionally use fw based ip discovery From Alex Deucher 840fe792a17069d99becbdd927822f175470cb3c in linux-6.12.y/6.12.36 80a0e828293389358f7db56adcdcb22b28df5e11 in mainline linux | |
|
cb82209c | 2025-07-11 00:19:30 | drm/scheduler: signal scheduled fence when kill job From Lin.Cao aefd0a935625165a6ca36d0258d2d053901555df in linux-6.12.y/6.12.36 471db2c2d4f80ee94225a1ef246e4f5011733e50 in mainline linux | |
|
f833c357 | 2025-07-11 00:16:58 | drm/amdgpu: seq64 memory unmap uses uninterruptible lock From Philip Yang 777580609d57861b8db3860c74142735d4c611e4 in linux-6.12.y/6.12.36 a359288ccb4dd8edb086e7de8fdf6e36f544c922 in mainline linux | |
|
88198963 | 2025-07-11 00:14:49 | amd/amdkfd: fix a kfd_process ref leak From Yifan Zhang 861204dce6ed0ff2ac1c1225129c11afe38aacfd in linux-6.12.y/6.12.36 90237b16ec1d7afa16e2173cc9a664377214cdd9 in mainline linux | |
 |
b42f7b38 | 2025-07-10 22:46:17 | Block interrupts while executing a command over the I2C bus as thei interrupt handler executes a command as well. This may result in a deadlock if an interrupt happens while executing a command. ok miod@ | |
 |
50c04e37 | 2025-07-10 14:27:43 | Some devices report a larger NCM sizes than wMaxSegmentSize. Rather limit them to UINT16_MAX than to wMaxSegmentSize. Found and OK by claudio@ | |
 |
767dd3b9 | 2025-07-10 13:59:27 | No arch uses ci_mutex_level in asm so remove it from genassym.cf OK kettenis@ | |
 |
d9289732 | 2025-07-10 13:37:14 | Add Microsoft Windows Dev Kit 2023 ok kettenis@ | |
 |
095ec576 | 2025-07-10 09:40:35 | sys/rtable.h: include sys/rwlock.h This header now depends on it. Unbreaks build on at least arm64. On amd64 this is pulled in via machine/cpu.h. discussed with/ok dlg | |
|
bdf12720 | 2025-07-10 05:28:13 | use the SMR api instead of SRPs in art/rtable code SRPs complicated this code massively, so moving it to SMR is almost like backing the SRP changes out. it becomes a lot more readable and similar to other ART implementations, which can't hurt. i modified how the heaps are traversed inside ART as well to try and minimise the pointer accesses as much as possible. the other big change is to walking over an ART. there's now a reentract and non-recursive iterator api which art_walk and rtable_walk are now built on top of. rtable_walk can basically do a foreach over the ART directly rather than having to pass a bunch of state into callbacks to apply with art_walk. we can still do that if we want, i just find this easier to deal with in rtable_walk directly. this part a part of a much bigger diff that was tested by hrvoje popovski, bluhm@, and matthieu@. mpi@ was keen ok jmatthew@ | |
|
148caabb | 2025-07-08 00:47:41 | remove unneeded includes; ok bluhm@ | |
|
9898077b | 2025-07-07 18:33:36 | Add an MI mechanism for creating an (unmapped) guard page between the PCB and the kernel stack. Switch amd64 over to this new mechanism. Also enable these guard pages on other 64-bit architectures with 4k pages (arm64, powerpc64, riscv64). ok miod@ | |
|
753322ee | 2025-07-07 18:06:35 | Fix various issues with backtraces such that we correctly display the instruction where we trapped to enter DDB and don't miss any frames. ok claudio@ | |
|
cded9c04 | 2025-07-07 07:09:05 | move ar_off out of struct art_root to r_off in struct rtable. this is the offset of an address in the relevant sockaddr, which rtable.c needs to know so it can pass the raw bytes of a network address to art for lookups. this part a part of a much bigger diff that was tested by hrvoje popovski, bluhm@, and matthieu@. ok jmatthew@ | |
|
debd0882 | 2025-07-07 06:51:45 | move route sourceaddr from struct art_root up to struct rtable. this part a part of a much bigger diff that was tested by hrvoje popovski, bluhm@, and matthieu@. ok jmatthew@ | |
|
177f10db | 2025-07-07 06:33:40 | create a struct rtable so we can take rtable bits out of art.c as a first step struct rtable sits between the rtmaps and the art_root, rather than have the rtmaps point directly at art_root. this the first part a part of a much bigger diff that was tested by hrvoje popovski, bluhm@, and matthieu@. ok jmatthew@ | |
|
6c5ade34 | 2025-07-07 02:28:50 | remove unneeded includes; ok bluhm@ | |
|
37d47b61 | 2025-07-07 00:55:15 | remove prototypes for removed functions | |
|
6cb1de41 | 2025-07-06 23:34:50 | remove pppoe_if_exists() prototype, function was never added | |
|
1a059e88 | 2025-07-06 12:22:31 | rework and try to unify ipi handlers. agintc and ampintc tried to allocate an interrupt per type of ipi, but could fall back to a combined ipi handler if needed. aplintc and bcm2836_intr always used a combined handler. i want to add another IPI in the future, so this diff sets it up to be as easy as possible by making all these drivers consistent around ipi handling. in this verson the type of IPI is encoded as a bit in a reason word, and then an IPI is sent to the target cpu. the ipi handler loops over the encoded bits and farms out handling to the more specific handlers as necessary. the nop ipi, which is the most common ipi, is optimised to avoid the atomic ops. it just needs the remote cpu to wake up, it doesnt rely on it runnning specific code, so we don't encode any bit in the reason word and just send the ipi. the handler special cases no bits set and returns immediately. testing and ok kettenis@ | |
|
78908cc8 | 2025-07-06 12:08:23 | Eliminate pointless double uuid endian switching when matching GPT partition types to filesystems to spoof. Kills last use of kern_uuid.c's uuid_dec_le() and most if not all invocations of kern_uuid.c's uuid_dec_be(). | |
|
4ad47e7c | 2025-07-06 11:07:09 | Someone had the bright idea to number pins starting from 1. Account for this issue by subtracting 1 from the pin number to find the correct registers. Make the sub-type per-pin, as on some of the PMICs it can vary. Add support for the PMIC variants found in x1e systems. ok patrick@ | |
|
ff609f0a | 2025-07-06 10:50:17 | Catch up with device tree binding changes. ok stsp@, phessler@ | |
 |
94a7e27d | 2025-07-06 02:11:58 | regen | |
|
b7dc1768 | 2025-07-06 02:10:13 | add another Lunar Lake id for an SMBus device ok jsg@ | |
|
54bc7180 | 2025-07-06 01:54:12 | remove duplicate defines | |
|
340de5d3 | 2025-07-05 23:44:49 | remove duplicate include of drm_drv.h | |
|
f2b2329d | 2025-07-05 22:54:53 | remove duplicate include of vmmvar.h | |
 |
2aea4a18 | 2025-07-05 19:03:02 | Prevent installing corrupted /bsd on relink errors Link with --no-mmap-output-file on LLD_ARCH to propagate failure, which defaults to mmap(2) and thus does not account for, e.g. running out of space in /usr/ or /tmp/. This causes targets like `newbsd', used by reorder_kernel and the installer, to install whatever ld.lld(1) produced despite errors before exiting zero. Seen by many, debugged by tb and semarie; undocumented flag found by jmatthew. "let's do it" deraadt | |
|
7ce78420 | 2025-07-05 16:36:30 | Remove unused GPT_UUID_* #define's. Move GPT_UUID_* #define's used when spoofing partitions to the location of their only invocation. | |
|
3c08d27e | 2025-07-05 09:24:37 | remove prototypes for removed functions | |
|
0168633b | 2025-07-05 04:25:43 | consistancy -> consistency | |
 |
c227130a | 2025-07-04 13:22:29 | Move declaration of __membar() outside of a !gcc4 conditional block. | |
|
663588fb | 2025-07-04 10:11:28 | spelling | |
|
e3bcd07f | 2025-07-04 09:34:48 | fix unintended double words in comments | |
|
b05c08e8 | 2025-07-04 04:36:25 | regen | |
|
81dc28f4 | 2025-07-04 04:35:58 | Actually add PCI Id for NetMos 9912 (previous had a copy/paste error) From Geoff Steckel, tweak/ok sthen | |
|
7b1b10eb | 2025-07-04 04:34:06 | Add PCI Id for NetMos 9912 From Geoff Steckel, tweak/ok sthen | |
|
3e6d8869 | 2025-07-04 04:31:48 | Teach puc(4) about my Nm9900 which attaches as puc0 at pci3 dev 0 function 0 "NetMos Nm9900" rev 0x00: ports: 16 com com4 at puc0 port 0 apic 2 int 16: st16650, 32 byte fifo ok deraadt kettenis sthen | |
|
f17b1aa7 | 2025-07-04 04:29:36 | regen | |
|
5397ce27 | 2025-07-04 04:29:15 | Add PCI Id for NetMos 9900 on my Dell Presicion's serial card ok deraadt kettenis sthen | |
|
b974508e | 2025-07-04 04:24:37 | pledge: allow IPV6_DONTFRAG optname in setsockopt() This is used by the latest version of quinn, firefox's QUIC implementation. ok deraadt | |
|
13b36e8a | 2025-07-03 21:06:51 | On Apple variants, enter DDB when a BREAK is detected. ok patrick@, jsg@, jca@ | |
|
2c6e3f42 | 2025-07-03 11:01:41 | set format bits for 8-bit unsigned linear encodings removed by mistake in rev 1.32 ok ratchov@ | |
|
d4ed460f | 2025-07-03 01:51:03 | drm/i915/pmu: Fix build error with GCOV and AutoFDO enabled From Tzung-Bi Shih fcc95344aa1c419d54f46c19011219363f05c593 in linux-6.12.y/6.12.35 a7137b1825b535eb7258b25beeb0d5425e0037d2 in mainline linux | |
|
ba1f95f8 | 2025-07-03 01:46:11 | drm/amdgpu: read back register after written for VCN v4.0.5 From David (Ming Qiang) Wu cef081c8231b52b4a652bbd022e03d80441118be in linux-6.12.y/6.12.35 ee7360fc27d6045510f8fe459b5649b2af27811a in mainline linux | |
|
e9a5d7e5 | 2025-07-02 21:28:46 | In AMD SEV-ES guest #VC trap asserts interrupts are disabled. Instead of disabling interrupts again, use kassert to verify that #VC trap stub in vector did not enable interrupts before calling vctrap(). Add a comment about the general workings of #VC handling. Prompted and provided by mlarkin@. from hshoexer@; OK mlarkin@ | |
 |
efc956da | 2025-07-02 16:44:40 | Move m_get(9) and m_split(9) calls out of mutex(9) protected section within the "while (((rcvstate & SS_RCVATMARK).." loop of somove(). The loop operates with local data so this is possible. The calls could sleep while somove() called from sosplice(). ok bluhm | |
|
51aaf22a | 2025-07-02 14:51:31 | Add an splassert(IPL_NONE) in fpu_kernel_enter() to catch unintended use of this function in interrupt context. ok claudio@, deraadt@ | |
|
e6ff7130 | 2025-07-02 14:50:05 | Rewrite m_getuio() to better align data in the mbufs Especially for datagram traffic data should be put at the end of a buffer that also holds at least max_hdr bytes extra space. Doing so should prevent the stack from calling m_prepend() later on. Also handle cluster allocations differently. If there is no cluster available just wait for memory and stop building stupidly long mbuf chains. In the very old times the number of clusters was very limited and so this was kind of needed but this has not been the case for a long time. Also most drivers will need to m_defrag() such a chain anyway which requires the same mbuf cluster to succeed. Joint work with dlg@ OK bluhm@ | |
|
1e2c81ff | 2025-07-02 14:36:56 | Make the mtx_owner pointer volatile in struct db_mutex else db_mtx_enter() may not work properly. OK kettenis@ miod@ | |
|
2327f17d | 2025-07-02 12:19:26 | run drm_managed cleanup from drm detach avoids a 'pool busy: still out' panic seen when radeondrm firmware is missing on non-efi installs reported by landry@ | |
|
ad175204 | 2025-07-02 05:44:46 | have route sourceaddr use RTF_GATEWAY to decide when to kick in. previously it used !RTF_HOST and !RTF_LLINFO. the intention with route sourceaddr was to use it except when a peer was on link. however, it is possible to have host routes (ie, RTF_HOST) via a gateway, which ended up not using the route sourceaddr when it should not have. by definition any route with RTF_GATEWAY set is not directly connected, so using it seems to better suit what route sourceaddr is doing. discussed with and ok claudio@ denis@ tested by denis@ | |
|
abfdae4f | 2025-07-01 11:10:36 | report the AMU feature ok kettenis@ | |
|
cffbe329 | 2025-06-30 14:19:20 | Drop IPL_SOFT/splsoft() ok jsg@ | |
|
76905433 | 2025-06-30 13:27:28 | AMD SEV-ES guest handles CPUID, IOIO, MSR during VC# trap. Fill in the handling of cpuid, in/out and rdmsr/wrmsr. For in/out and rdmsr/wrmsr decode the actual instructions. Use the GHCB to request emulation of these instructions from vmm(4). With this commit, SEV-ES enabled guest are able to run on vmm(4)/vmd(8). from hshoexer@; OK mlarkin@ | |
 |
af62f95d | 2025-06-30 13:01:10 | BACKOUT: vport/veb(4): Use hardware checksum offload. Due to some issues pointed out by bluhm, I have to backout this diff. | |
|
3ae74042 | 2025-06-30 12:43:22 | Unlock IPCTL_DEFTTL case of ip_sysctl(). Read-only access at initialization time of interfaces, PCBs and IP packets. ok bluhm | |
|
fee98b1b | 2025-06-30 11:39:50 | allowpowerdown extern no longer required | |
|
1e731c47 | 2025-06-30 04:40:03 | unify style of character device tests; no functional change feedback and ok deraadt@ | |
|
344c04d6 | 2025-06-29 19:32:08 | Remove unused routines. | |
|
8a937f11 | 2025-06-29 15:55:21 | Remove whitespace at EOL (except in licence blocks), replace sets of spaces with tabs. ok jsg@ | |
|
c5891aa6 | 2025-06-29 13:48:34 | vport/veb(4): Use hardware checksum offload. ok dlg@ | |
|
948e6e0c | 2025-06-29 11:24:10 | Avoid NULL pointer dereference when TIOCUCNTL ioctl is called with a non-file argument. feedback and ok deraadt@ | |
|
ca2e87ba | 2025-06-29 09:31:47 | Install final #VC trap handler for AMD SEV-ES guests. As OUT instructions are used to ack, mask and unmask interrupts in the PIC, they will raise #VC exceptions while handling interrupts. Most exceptions re-enable interrupts as soon as the stack frame is setup. However, in case of #VC we must disable interrupts during the whole exception handling. Otherwise we would get nested IRQs of the same PSL. Therefore do not use the TRAP() macro but rewrite the entry stub without enabling interrupts and jump to the "common" code as soon as possible. from hshoexer@; OK mlarkin@ | |
|
6036672b | 2025-06-29 00:33:46 | make the argument to ether_ntoa const. this drives me nuts when i want to print something out of what's already const. casting it works, but feels gross. ok guenther@ tb@ deraadt@ enh says this is already in bionic/glibc/musl | |
|
31aef2a8 | 2025-06-28 16:04:09 | Convert the last few K&R function definitions to ANSI. NFC ok jsg@ | |
|
b3db3559 | 2025-06-28 13:24:21 | Convert the last few K&R function definitions to ANSI. NFC ok jsg@ | |
 |
5f143f75 | 2025-06-28 12:43:17 | iwx(4) in RAMDISK for riscv64. From Shawn Chiou who tested on AX200. | |
|
0ec6f054 | 2025-06-28 11:34:21 | No need to include <iommureg.h> in these files anymore. | |
|
ec0bf924 | 2025-06-28 11:34:01 | Remove unused interrupt macros. | |
|
7e620f5f | 2025-06-28 11:33:45 | Move interrupt-related defines from iommureg.h to intreg.h. NFC | |
|
f49e0588 | 2025-06-28 11:33:12 | Remove dead code. | |
|
474681e2 | 2025-06-28 04:21:15 | Enable iwx(4) from Shawn Chiou who tested on AX200. | |
|
320bfc24 | 2025-06-27 17:23:49 | Allocate GHCB and handle #VC trap early in AMD SEV-ES guest. When the kernel runs as SEV-ES guest, claim two pages to be used for GHCB communication with vmm(4). As the guest wants to share these two pages with vmm(4) host, re-map them as plain text. Do this as soon as we reach long mode. Arriving in init_x86_64(), reset the IDT used during locore0 and install a new #VC trap handler. This is the actual handler the kernel will use from now on. However, as we are not fully up and running yet, use a simple and kernel-only entry stub for the #VC exception. At the end of early bootstrap, init_x86_64() will install the actual and final IDT. For now this will reset and disable the #VC handler. Therefore, SEV-ES enabled guest are not working yet. The #VC handler vctrap() is also hooked into kerntrap() and usertrap(). Communication with vmm(4) using the GHCB protocol is also ready. However, the actual handling of cpuid, in/out, wrmsr/rdmsr is left out for now. Again, SEV-ES enabled guest are not working yet. from hshoexer@; OK mlarkin@ | |
|
81e8e60e | 2025-06-27 16:18:10 | Use SoftLRO in ice(4), but default off. We don't use the ptype decode, because we just want to use SoftLRO in cases of TCP with just plain IPv4 and IPv6 packets. Tested by bluhm@ ok bluhm@ | |
|
1800221d | 2025-06-27 09:25:57 | vport/veb(4): Use VLAN Hardware Tagging. ok dlg | |
|
2a3a2a0f | 2025-06-26 21:46:40 | Fix TCP netstat counter. In my previous commit I forgot an ASSIGN() in tcp_sysctl_tcpstat(). Then the counter index was not incremented. This caused that values in netstat -s were of by one in position with the description. reported and OK jan@ | |
|
2c286071 | 2025-06-26 20:43:41 | Factor code responsible for flipping bits in the auxio registers. Only include the floppy routine if the floppy driver is configured. | |
|
da3b121c | 2025-06-26 20:39:34 | Don't bother mapping auxio registers we don't care about. |