skiboot-6.4¶
skiboot v6.4 was released on Tuesday July 16th 2019. It is the first release of skiboot 6.4, which becomes the new stable release of skiboot following the 6.3 release, first released May 3rd 2019.
Skiboot 6.4 will mark the basis for op-build v2.4.
skiboot v6.4 contains all bug fixes as of skiboot-6.0.20, and skiboot-6.3.2 (the currently maintained stable releases).
For how the skiboot stable releases work, see Skiboot stable tree rules and releases for details.
Over skiboot 6.3, we have the following changes:
New features¶
Since skiboot v6.4-rc1:
npu2-opencapi: Add opencapi support on ZZ
This patch adds opencapi support on ZZ. It hard-codes the required device tree entries for the NPU and links. The alternative was to use HDAT, but it somehow proved too painful to do.
The new device tree entries activate the npu2 init code on ZZ. On systems with no opencapi adapters, it should go unnoticed, as presence detection will skip link training.
Since skiboot v6.3:
platforms/nicole: Add new platform
The platform is a new platform from YADRO, it’s a storage controller for TATLIN server. It’s Based on IBM Romulus reference design (POWER9).
platform/zz: Add new platform type
We have new platform type under ZZ. Lets add them. With this fix
nvram: Flag dangerous NVRAM options
Most nvram options used by skiboot are just for debug or testing for regressions. They should never be used long term.
We’ve hit a number of issues in testing and the field where nvram options have been set “temporarily” but haven’t been properly cleared after, resulting in crashes or real bugs being masked.
This patch marks most nvram options used by skiboot as dangerous and prints a chicken to remind users of the problem.
hw/phb3: Add verbose EEH output
Add support for the pci-eeh-verbose NVRAM flag on PHB3. We’ve had this on PHB4 since forever and it has proven very useful when debugging EEH issues. When testing changes to the Linux kernel’s EEH implementation it’s fairly common for the kernel to crash before printing the EEH log so it’s helpful to have it in the OPAL log where it can be dumped from XMON.
Note that unlike PHB4 we do not enable verbose mode by default. The nvram option must be used to explicitly enable it.
Experimental support for building without FSP code
Now, with CONFIG_FSP=0/1 we have:
1.6M/1.4M skiboot.lid
323K/375K skiboot.lid.xz
doc: travis-ci deploy docs!
Documentation is now automatically deployed if you configure Travis CI appropriately (we have done this for the open-power branch of skiboot)
Big OPAL API Documentation improvement
A lot more OPAL API calls are now (at least somewhat) documented.
opal/hmi: Report NPU2 checkstop reason
The NPU2 is currently not passing any information to linux to explain the cause of an HMI. NPU2 has three Fault Isolation Registers and over 30 of those FIR bits are configured to raise an HMI by default. We won’t be able to fit all possible state in the 32-bit xstop_reason field of the HMI event, but we can still try to encode up to 4 HMI reasons.
opal-msg: Enhance opal-get-msg API
Linux uses OPAL_GET_MSG API to get OPAL messages. This interface supports upto 8 params (64 bytes). We have a requirement to send bigger data to Linux. This patch enhances OPAL to send bigger data to Linux.
Linux will use “opal-msg-size” device tree property to allocate memory for OPAL messages (previous patch increased “opal-msg-size” to 64K).
Replaced reserved field in “struct opal_msg” with size. So that Linux side opal_get_msg user can detect actual data size.
If buffer size < actual message size, then opal_get_msg will copy partial data and return OPAL_PARTIAL to Linux.
Add new variable “extended” to “opal_msg_entry” structure to keep track of messages that has more than 64byte data. We will allocate separate memory for these messages and once kernel consumes message we will release that memory.
core/opal: Increase opal-msg-size size
Kernel will use opal-msg-size property to allocate memory for opal_msg. We want to send bigger data from OPAL to kernel. Hence increase opal-msg-size to 64K.
hw/npu2-opencapi: Add initial support for allocating OpenCAPI LPC memory
Lowest Point of Coherency (LPC) memory allows the host to access memory on an OpenCAPI device.
Define 2 OPAL calls, OPAL_NPU_MEM_ALLOC and OPAL_NPU_MEM_RELEASE, for assigning and clearing the memory BAR. (We try to avoid using the term “LPC” to avoid confusion with Low Pin Count.)
At present, we use a fixed location in the address space, which means we are restricted to a single range of 4TB, on a single OpenCAPI device per chip. In future, we’ll use some chip ID extension magic to give us more space, and some sort of allocator to assign ranges to more than one device.
core/fast-reboot: Add im-feeling-lucky option
Fast reboot gets disabled for a number of reasons e.g. the availability of nvlink. However this doesn’t actually affect the ability to perform fast reboot if no nvlink device is actually present.
Add a nvram option for fast-reset where if it’s set to “im-feeling-lucky” then perform the fast-reboot irrespective of if it’s previously been disabled.
platforms/astbmc: Check for SBE validation step
On some POWER8 astbmc systems an update to the SBE requires pausing at runtime to ensure integrity of the SBE. If this is required the BMC will set a chassis boot option IPMI flag using the OEM parameter 0x62. If Skiboot sees this flag is set it waits until the SBE update is complete and the flag is cleared.
Unfortunately the mystery operation that validates the SBE also leaves it in a bad state and unable to be used for timer operations. To workaround this the flag is checked as soon as possible (ie. when IPMI and the console are set up), and once complete the system is rebooted.
Add P9 DIO interrupt support
On P9 there are GPIO port 0, 1, 2 for GPIO interrupt, and DIO interrupt is used to handle the interrupts.
Add support to the DIO interrupts:
Add dio_interrupt_register(chip, port, callback) to register the interrupt
Add dio_interrupt_deregister(chip, port, callback) to deregister;
When interrupt on the port occurs, callback is invoked, and the interrupt status is cleared.
Removed features¶
Since skiboot v6.3:
pci/iov: Remove skiboot VF tracking
This feature was added a few years ago in response to a request to make the MaxPayloadSize (MPS) field of a Virtual Function match the MPS of the Physical Function that hosts it.
The SR-IOV specification states the the MPS field of the VF is “ResvP”. This indicates the VF will use whatever MPS is configured on the PF and that the field should be treated as a reserved field in the config space of the VF. In other words, a SR-IOV spec compliant VF should always return zero in the MPS field. Adding hacks in OPAL to make it non-zero is… misguided at best.
Additionally, there is a bug in the way pci_device structures are handled by VFs that results in a crash on fast-reboot that occurs if VFs are enabled and then disabled prior to rebooting. This patch fixes the bug by removing the code entirely. This patch has no impact on SR-IOV support on the host operating system.
Remove POWER7 and POWER7+ support
It’s been a good long while since either OPAL POWER7 user touched a machine, and even longer since they’d have been okay using an old version rather than tracking master.
There’s also been no testing of OPAL on POWER7 systems for an awfully long time, so it’s pretty safe to assume that it’s very much bitrotted.
It also saves a whole 14kb of xz compressed payload space.
Remove remnants of OPAL_PCI_GET_PHB_DIAG_DATA
Never present in a public OPAL release, and only kernels prior to 3.11 would ever attempt to call it.
Remove unused OPAL_GET_XIVE_SOURCE
While this call was technically implemented by skiboot, no code has ever called it, and it was only ever implemented for the p7ioc-phb back-end (i.e. POWER7). Since this call was unused in Linux, and that POWER7 with OPAL was only ever available internally, so it should be safe to remove the call.
Remove unused OPAL_PCI_GET_XIVE_REISSUE and OPAL_PCI_SET_XIVE_REISSUE
These seem to be remnants of one of the OPAL incarnations prior to OPALv3. These calls have never been implemented in skiboot, and never used by an upstream kernel (nor a PowerKVM kernel).
It’s rather safe to just document them as never existing.
Remove never implemented OPAL_PCI_SET_PHB_TABLE_MEMORY and document why
Not ever used by upstream linux or PowerKVM tree. Never implemented in skiboot (not even in ancient internal only tree).
So, it’s incredibly safe to remove.
Remove unused OPAL_PCI_EEH_FREEZE_STATUS2
This call was introduced all the way back at the end of 2012, before OPAL was public. The #define for the OPAL call was introduced to the Linux kernel in June 2013, and the call was never used in any kernel tree ever (as far as we can find).
Thus, it’s quite safe to remove this completely unused and completely untested OPAL call.
Document the long removed OPAL_REGISTER_OPAL_EXCEPTION_HANDLER call
I’m pretty sure this was removed in one of our first ever service packs.
Remove last remnants of OPAL_PCI_SET_PHB_TCE_MEMORY and OPAL_PCI_SET_HUB_TCE_MEMORY
Since we have not supported p5ioc systems since skiboot 5.2, it’s pretty safe to just wholesale remove these OPAL calls now.
Remove remnants of OPAL_PCI_SET_PHB_TCE_MEMORY
There’s no reason we need remnants hanging around that aren’t used, so remove them and save a handful of bytes at runtime.
Simultaneously, document the OPAL call removal.
Secure and Trusted Boot¶
Since skiboot v6.3:
trustedboot: Change PCR and event_type for the skiboot events
The existing skiboot events are being logged as EV_ACTION, however, the TCG PC Client spec says that EV_ACTION events should have one of the pre-defined strings in the event field recorded in the event log. For instance:
“Calling Ready to Boot”,
“Entering ROM Based Setup”,
“User Password Entered”, and
“Start Option ROM Scan.
None of the EV_ACTION pre-defined strings are applicable to the existing skiboot events. Based on recent discussions with other POWER teams, this patch proposes a convention on what PCR and event types should be used for skiboot events. This also changes the skiboot source code to follow the convention.
The TCG PC Client spec defines several event types, other than EV_ACTION. However, many of them are specific to UEFI events and some others are related to platform or CRTM events, which is more applicable to hostboot events.
Currently, most of the hostboot events are extended to PCR[0,1] and logged as either EV_PLATFORM_CONFIG_FLAGS, EV_S_CRTM_CONTENTS or EV_POST_CODE. The “Node Id” and “PAYLOAD” events, though, are extended to PCR[4,5,6] and logged as EV_COMPACT_HASH.
For the lack of an event type that fits the specific purpose, EV_COMPACT_HASH seems to be the most adequate one due to its flexibility. According to the TCG PC Client spec:
May be used for any PCR except 0, 1, 2 and 3.
The event field may be informative or may be hashed to generate the digest field, depending on the component recording the event.
Additionally, the PCR[4,5] seem to be the most adequate PCRs. They would be used for skiboot and some skiroot events. According to the TCG PC Client, PCR[4] is intended to represent the entity that manages the transition between the pre-OS and OS-present state of the platform. PCR[4], along with PCR[5], identifies the initial OS loader.
In summary, for skiboot events:
Events that represents data should be extended to PCR 4.
Events that represents config should be extended to PCR 5.
For the lack of an event type that fits the specific purpose, both data and config events should be logged as EV_COMPACT_HASH.
Sensors¶
Since skiboot v6.3:
occ-sensors: Check if OCC is reset while reading inband sensors
OCC may not be able to mark the sensor buffer as invalid while going down RESET. If OCC never comes back we will continue to read the stale sensor data. So verify if OCC is reset while reading the sensor values and propagate the appropriate error.
IPMI¶
Since skiboot v6.3:
ipmi: ensure forward progress on ipmi_queue_msg_sync()
BT responses are handled using a timer doing the polling. To hope to get an answer to an IPMI synchronous message, the timer needs to run.
We can’t just check all timers though as there may be a timer that wants a lock that’s held by a code path calling ipmi_queue_msg_sync(), and if we did enforce that as a requirement, it’s a pretty subtle API that is asking to be broken.
So, if we just run a poll function to crank anything that the IPMI backend needs, then we should be fine.
This issue shows up very quickly under QEMU when loading the first flash resource with the IPMI HIOMAP backend.
NPU2¶
Since skiboot v6.4-rc1:
witherspoon: Add nvlink peers in finalise_dt()
This information is consumed by Linux so it needs to be in the DT. Move it to finalise_dt().
Since skiboot v6.3:
npu2: Increase timeout for L2/L3 cache purging
On NVLink2 bridge reset, we purge all L2/L3 caches in the system. This is an asynchronous operation, we have a 2ms timeout here. There are reports that this is not enough and “PURGE L3 on core xxx timed out” messages appear (for the reference: on the test setup this takes 280us..780us).
This defines the timeout as a macro and changes this from 2ms to 20ms.
This adds a tracepoint to tell how long it took to purge all the caches.
npu2: Purge cache when resetting a GPU
After putting all a GPU’s links in reset, do a cache purge in case we have CPU cache lines belonging to the now-unaccessible GPU memory.
npu2-opencapi: Mask 2 XSL errors
Commit f8dfd699f584 (“hw/npu2: Setup an error interrupt on some opencapi FIRs”) converted some FIR bits default action from system checkstop to raising an error interrupt. For 2 XSL error events that can be triggered by a misbehaving AFU, the error interrupt is raised twice, once for each link (the XSL logic in the NPU is shared between 2 links). So a badly behaving AFU could impact another, unsuspecting opencapi adapter.
It doesn’t look good and it turns out we can do better. We can mask those 2 XSL errors. The error will also be picked up by the OTL logic, which is per link. So we’ll still get an error interrupt, but only on the relevant link, and the other opencapi adapter can stay functional.
npu2: Clear fence state for a brick being reset
Resetting a GPU before resetting an NVLink leads to occasional HMIs which fence some bricks and prevent the “reset_ntl” procedure from succeeding at the “reset_ntl_release” step - the host system requires reboot; there may be other cases like this as well.
This adds clearing of the fence bit in NPU.MISC.FENCE_STATE for the NVLink which we are about to reset.
npu2: Fix clearing the FIR bits
FIR registers are SCOM-only so they cannot be accesses with the indirect write, and yet we use SCOM-based addresses for these; fix this.
npu2: Reset NVLinks when resetting a GPU
Resetting a V100 GPU brings its NVLinks down and if an NPU tries using those, an HMI occurs. We were lucky not to observe this as the bare metal does not normally reset a GPU and when passed through, GPUs are usually before NPUs in QEMU command line or Libvirt XML and because of that NPUs are naturally reset first. However simple change of the device order brings HMIs.
This defines a bus control filter for a PCI slot with a GPU with NVLinks so when the host system issues secondary bus reset to the slot, it resets associated NVLinks.
npu2: Reset PID wildcard and refcounter when mapped to LPID
Since 105d80f85b “npu2: Use unfiltered mode in XTS tables” we do not register every PID in the XTS table so the table has one entry per LPID. Then we added a reference counter to keep track of the entry use when switching GPU between the host and guest systems (the “Fixes:” tag below).
The POWERNV platform setup creates such entries and references them at the boot time when initializing IOMMUs and only removes it when a GPU is passed through to a guest. This creates a problem as POWERNV boots via kexec and no defererencing happens; the XTS table state remains undefined. So when the host kernel boots, skiboot thinks there are valid XTS entries and does not update the XTS table which breaks ATS.
This adds the reference counter and the XTS entry reset when a GPU is assigned to LPID and we cannot rely on the kernel to clean that up.
PHB4¶
Since skiboot v6.3:
hw/phb4: Make phb4_training_trace() more general
phb4_training_trace() is used to monitor the Link Training Status State Machine (LTSSM) of the PHB’s data link layer. Currently it is only used to observe the LTSSM while bringing up the link, but sometimes it’s useful to see what’s occurring in other situations (e.g. link disable, or secondary bus reset). This patch renames it to phb4_link_trace() and allows the target LTSSM state and a flexible timeout to help in these situations.
hw/phb4: Make pci-tracing print at PR_NOTICE
When pci-tracing is enabled we print each trace status message and the final trace status at PR_ERROR. The final status messages are similar to those printed when we fail to train in the non-pci-tracing path and this has resulted in spurious op-test failures.
This patch reduces the log-level of the tracing message to PR_NOTICE so they’re not accidently interpreted as actual error messages. PR_NOTICE messages are still printed to the console during boot.
hw/phb4: Use read/write_reg in assert_perst
While the PHB is fenced we can’t use the MMIO interface to access PHB registers. While processing a complete reset we inject a PHB fence to isolate the PHB from the rest of the system because the PHB won’t respond to MMIOs from the rest of the system while being reset.
We assert PERST after the fence has been erected which requires us to use the XSCOM indirect interface to access the PHB registers rather than the MMIO interface. Previously we did that when asserting PERST in the CRESET path. However in b8b4c79d4419 (“hw/phb4: Factor out PERST control”). This was re-written to use the raw in_be64() accessor. This means that CRESET would not be asserted in the reset path. On some Mellanox cards this would prevent them from re-loading their firmware when the system was fast-reset.
This patch fixes the problem by replacing the raw {in|out}_be64() accessors with the phb4_{read|write}_reg() functions.
hw/phb4: Assert Link Disable bit after ETU init
The cursed RAID card in ozrom1 has a bug where it ignores PERST being asserted. The PCIe Base spec is a little vague about what happens while PERST is asserted, but it does clearly specify that when PERST is de-asserted the Link Training and Status State Machine (LTSSM) of a device should return to the initial state (Detect) defined in the spec and the link training process should restart.
This bug was worked around in 9078f8268922 (“phb4: Delay training till after PERST is deasserted”) by setting the link disable bit at the start of the FRESET process and clearing it after PERST was de-asserted. Although this fixed the bug, the patch offered no explaination of why the fix worked.
In b8b4c79d4419 (“hw/phb4: Factor out PERST control”) the link disable workaround was moved into phb4_assert_perst(). This is called always in the CRESET case, but a following patch resulted in assert_perst() not being called if phb4_freset() was entered following a CRESET since p->skip_perst was set in the CRESET handler. This is bad since a side-effect of the CRESET is that the Link Disable bit is cleared.
This, combined with the RAID card ignoring PERST results in the PCIe link being trained by the PHB while we’re waiting out the 100ms ETU reset time. If we hack skiboot to print a DLP trace after returning from phb4_hw_init() we get:
PHB#0001[0:1]: Initialization complete PHB#0001[0:1]: TRACE:0x0000102101000000 0ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000001101000000 23ms GEN1:x16:detect PHB#0001[0:1]: TRACE:0x0000102101000000 23ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000183101000000 29ms training GEN1:x16:config PHB#0001[0:1]: TRACE:0x00001c5881000000 30ms training GEN1:x08:recovery PHB#0001[0:1]: TRACE:0x00001c5883000000 30ms training GEN3:x08:recovery PHB#0001[0:1]: TRACE:0x0000144883000000 33ms presence GEN3:x08:L0 PHB#0001[0:1]: TRACE:0x0000154883000000 33ms trained GEN3:x08:L0 PHB#0001[0:1]: CRESET: wait_time = 100 PHB#0001[0:1]: FRESET: Starts PHB#0001[0:1]: FRESET: Prepare for link down PHB#0001[0:1]: FRESET: Assert skipped PHB#0001[0:1]: FRESET: Deassert PHB#0001[0:1]: TRACE:0x0000154883000000 0ms trained GEN3:x08:L0 PHB#0001[0:1]: TRACE: Reached target state PHB#0001[0:1]: LINK: Start polling PHB#0001[0:1]: LINK: Electrical link detected PHB#0001[0:1]: LINK: Link is up PHB#0001[0:1]: LINK: Went down waiting for stabilty PHB#0001[0:1]: LINK: DLP train control: 0x0000105101000000 PHB#0001[0:1]: CRESET: Starts
What has happened here is that the link is trained to 8x Gen3 33ms after we return from phb4_init_hw(), and before we’ve waitined to 100ms that we normally wait after re-initialising the ETU. When we “deassert” PERST later on in the FRESET handler the link in L0 (normal) state. At this point we try to read from the Vendor/Device ID register to verify that the link is stable and immediately get a PHB fence due to a PCIe Completion Timeout. Skiboot attempts to recover by doing another CRESET, but this will encounter the same issue.
This patch fixes the problem by setting the Link Disable bit (by calling phb4_assert_perst()) immediately after we return from phb4_init_hw(). This prevents the link from being trained while PERST is asserted which seems to avoid the Completion Timeout. With the patch applied we get:
PHB#0001[0:1]: Initialization complete PHB#0001[0:1]: TRACE:0x0000102101000000 0ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000001101000000 23ms GEN1:x16:detect PHB#0001[0:1]: TRACE:0x0000102101000000 23ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000909101000000 29ms presence GEN1:x16:disabled PHB#0001[0:1]: CRESET: wait_time = 100 PHB#0001[0:1]: FRESET: Starts PHB#0001[0:1]: FRESET: Prepare for link down PHB#0001[0:1]: FRESET: Assert skipped PHB#0001[0:1]: FRESET: Deassert PHB#0001[0:1]: TRACE:0x0000001101000000 0ms GEN1:x16:detect PHB#0001[0:1]: TRACE:0x0000102101000000 0ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000001101000000 24ms GEN1:x16:detect PHB#0001[0:1]: TRACE:0x0000102101000000 36ms presence GEN1:x16:polling PHB#0001[0:1]: TRACE:0x0000183101000000 97ms training GEN1:x16:config PHB#0001[0:1]: TRACE:0x00001c5881000000 97ms training GEN1:x08:recovery PHB#0001[0:1]: TRACE:0x00001c5883000000 97ms training GEN3:x08:recovery PHB#0001[0:1]: TRACE:0x0000144883000000 99ms presence GEN3:x08:L0 PHB#0001[0:1]: TRACE: Reached target state PHB#0001[0:1]: LINK: Start polling PHB#0001[0:1]: LINK: Electrical link detected PHB#0001[0:1]: LINK: Link is up PHB#0001[0:1]: LINK: Link is stable PHB#0001[0:1]: LINK: Card [9005:028c] Optimal Retry:disabled PHB#0001[0:1]: LINK: Speed Train:GEN3 PHB:GEN4 DEV:GEN3 PHB#0001[0:1]: LINK: Width Train:x08 PHB:x08 DEV:x08 PHB#0001[0:1]: LINK: RX Errors Now:0 Max:8 Lane:0x0000
Simulators¶
Since skiboot v6.3:
external/mambo: Bump default POWER9 to Nimbus DD2.3
external/mambo: fix tcl startup code for mambo bogus net (repost)
This fixes a couple issues with external/mambo/skiboot.tcl so I can use the mambo bogus net.
newer distros (ubuntu 18.04) allow tap device to have a user specified name instead of just tapN so we need to pass in a name not a number.
need some kind of default for net_mac, and need the mconfig for it to be set from an env var.
skiboot.tcl: Add option to wait for GDB server connection
Add an environment variable which makes Mambo wait for a connection from gdb prior to starting simulation.
mambo: Integrate addr2line into backtrace command
Gives nice output like this:
systemsim % bt pc: 0xC0000000002BF3D4 _savegpr0_28+0x0 lr: 0xC00000000004E0F4 opal_call+0x10 stack:0x000000000041FAE0 0xC00000000004F054 opal_check_token+0x20 stack:0x000000000041FB50 0xC0000000000500CC __opal_flush_console+0x88 stack:0x000000000041FBD0 0xC000000000050BF8 opal_flush_console+0x24 stack:0x000000000041FC00 0xC0000000001F9510 udbg_opal_putc+0x88 stack:0x000000000041FC40 0xC000000000020E78 udbg_write+0x7c stack:0x000000000041FC80 0xC0000000000B1C44 console_unlock+0x47c stack:0x000000000041FD80 0xC0000000000B2424 register_console+0x320 stack:0x000000000041FE10 0xC0000000003A5328 register_early_udbg_console+0x98 stack:0x000000000041FE80 0xC0000000003A4F14 setup_arch+0x68 stack:0x000000000041FEF0 0xC0000000003A0880 start_kernel+0x74 stack:0x000000000041FF90 0xC00000000000AC60 start_here_common+0x1c
mambo: Add addr2func for symbol resolution
If you supply a VMLINUX_MAP/SKIBOOT_MAP/USER_MAP addr2func can guess at your symbol name. i.e.
systemsim % p pc 0xC0000000002A68F8 systemsim % addr2func [p pc] fdt_offset_ptr+0x78
lpc-port80h: Don’t write port 80h when running under Simics
Simics doesn’t model LPC port 80h. Writing to it terminates the simulation due to an invalid LPC memory access. This patch adds a check to ensure port 80h isn’t accessed if we are running under Simics.
device-tree: speed up fdt building on slow simulators
Trade size for speed and avoid de-duplicating strings in the fdt. This costs about 2kB in fdt size, and saves about 8 million instructions (almost half of all instructions) booting skiboot in mambo.
fast-reboot:: skip read-only memory checksum for slow simulators
Skip the fast reboot checksum, which costs about 4 million cycles booting skiboot in mambo.
nx: remove check on the “qemu, powernv” property
commit 95f7b3b9698b (“nx: Don’t abort on missing NX when using a QEMU machine”) introduced a check on the property “qemu,powernv” to skip NX initialization when running under a QEMU machine.
The QEMU platforms now expose a QUIRK_NO_RNG in the chip. Testing the “qemu,powernv” property is not necessary anymore.
plat/qemu: add a POWER8 and POWER9 platform
These new QEMU platforms have characteristics closer to real OpenPOWER systems that we use today and define a different BMC depending on the CPU type. New platform properties are introduced for each, “qemu,powernv8”, “qemu,powernv9” and these should be compatible with existing QEMUs which only expose the “qemu,powernv” property
libc/string: speed up common string functions
Use compiler builtins for the string functions, and compile the libc/string/ directory with -O2.
This reduces instructions booting skiboot in mambo by 2.9 million in slow-sim mode, or 3.8 in normal mode, for less than 1kB image size increase.
This can result in the compiler warning more cases of string function problems.
external/mambo: Add an option to exit Mambo when the system is shutdown
Automatically exiting can be convenient for scripting. Will also exit due to a HW crash (eg. unhandled exception).
VESNIN platform¶
Since skiboot v6.3:
platforms/vesnin: PCI inventory via IPMI OEM
Replace raw protocol with OEM message supported by OpenBMC’s IPMI plugins.
BMC-side implementation (IPMI plug-in): https://github.com/YADRO-KNS/phosphor-pci-inventory
Utilities¶
Since skiboot v6.3:
opal-gard: Account for ECC size when clearing partition
When ‘opal-gard clear all’ is run, it works by erasing the GUARD then using blockevel_smart_write() to write nothing to the partition. This second write call is needed because we rely on libflash to set the ECC bits appropriately when the partition contained ECCed data.
The API for this is a little odd with the caller specifying how much actual data to write, and libflash writing size + size/8 bytes since there is one additional ECC byte for every eight bytes of data.
We currently do not account for the extra space consumed by the ECC data in reset_partition() which is used to handle the ‘clear all’ command. Which results in the paritition following the GUARD partition being partially overwritten when the command is used. This patch fixes the problem by reducing the length we would normally write by the number of ECC bytes required.
Build and debugging¶
Since skiboot v6.3:
Disable -Waddress-of-packed-member for GCC9
We throw a bunch of errors in errorlog code otherwise, which we should fix, but we don’t have to yet.
Fix a lot of sparse warnings
With new GCC comes larger GCOV binaries
So we need to change our heap size to make more room for data/bss without having to change where the console is or have more fun moving things about.
Intentionally discard fini_array sections
Produced in a SKIBOOT_GCOV=1 build, and never called by skiboot.
external/trace: Add follow option to dump_trace
When monitoring traces, an option like the tail command’s ‘-f’ (follow) is very useful. This option continues to append to the output as more data arrives. Add an ‘-f’ option to allow dump_trace to operate similarly.
Tail also provides a ‘-s’ (sleep time) option that accompanies ‘-f’. This controls how often new input will be polled. Add a ‘-s’ option that will make dump_trace sleep for N milliseconds before checking for new input.
external/trace: Add support for dumping multiple buffers
dump_trace only can dump one trace buffer at a time. It would be handy to be able to dump multiple buffers and to see the entries from these buffers displayed in correct timestamp order. Each trace buffer is already sorted by timestamp so use a heap to implement an efficient k-way merge. Use the CCAN heap to implement this sort. However the CCAN heap does not have a ‘heap_replace’ operation. We need to ‘heap_pop’ then ‘heap_push’ to replace the root which means rebalancing twice instead of once.
external/trace: mmap trace buffers in dump_trace
The current lseek/read approach used in dump_trace does not correctly handle certain aspects of the buffers. It does not use the start and end position that is part of the buffer so it will not begin from the correct location. It does not move back to the beginning of the trace buffer file as the buffer wraps around. It also does not handle the overflow case of the writer overwriting when the reader is up to.
Mmap the trace buffer file so that the existing reading functions in extra/trace.c can be used. These functions already handle the cases of wrapping and overflow. This reduces code duplication and uses functions that are already unit tested. However this requires a kernel where the trace buffer sysfs nodes are able to be mmaped (see https://patchwork.ozlabs.org/patch/1056786/)
core/trace: Export trace buffers to sysfs
Every property in the device-tree under /ibm,opal/firmware/exports has a sysfs node created in /firmware/opal/exports. Add properties with the physical address and size for each trace buffer so they are exported.
core/trace: Add pir number to debug_descriptor
The names given to the trace buffers when exported to sysfs should show what cpu they are associated with to make it easier to understand there output. The debug_descriptor currently stores the address and length of each trace buffer and this is used for adding properties to the device tree. Extend debug_descriptor to include a cpu associated with each trace. This will be used for creating properties in the device-tree under /ibm,opal/firmware/exports/.
core/trace: Change trace buffer size
We want to be able to mmap the trace buffers to be used by the dump_trace tool. As mmaping is done in terms of pages it makes sense that the size of the trace buffers should be page aligned. This is slightly complicated by the space taken up by the header at the beginning of the trace and the room left for an extra trace entry at the end of the buffer. Change the size of the buffer itself so that the entire trace buffer size will be page aligned.
core/trace: Change buffer alignment from 4K to 64K
We want to be able to mmap the trace buffers to be used by the dump_trace tool. This means that the trace bufferes must be page aligned. Currently they are aligned to 4K. Most power systems have a 64K page size. On systems with a 4K page size, 64K aligned will still be page aligned. Change the allocation of the trace buffers to be 64K aligned.
The trace_info struct that contains the trace buffer is actually what is allocated aligned memory. This means the trace buffer itself is not actually aligned and this is the address that is currently exposed through sysfs. To get around this change the address that is exposed to sysfs to be the trace_info struct. This means the lock in trace_info is now visible too.
external/trace: Use correct width integer byte swapping
The trace_repeat struct uses be16 for storing the number of repeats. Currently be32_to_cpu conversion is used to display this member. This produces an incorrect value. Use be16_to_cpu instead.
core/trace: Put boot_tracebuf in correct location.
A position for the boot_tracebuf is allocated in skiboot.lds.S. However, without a __section attribute the boot trace buffer is not placed in the correct location, meaning that it also will not be correctly aligned. Add the __section attribute to ensure it will be placed in its allocated position.
core/lock: Add debug options to store backtrace of where lock was taken
Contrary to popular belief, skiboot developers are imperfect and occasionally write locking bugs. When we exit skiboot, we check if we’re still holding any locks, and if so, we print an error with a list of the locks currently held and the locations where they were taken.
However, this only tells us the location where lock() was called, which may not be enough to work out what’s going on. To give us more to go on with, we can store backtrace data in the lock and print that out when we unexpectedly still hold locks.
Because the backtrace data is rather big, we only enable this if DEBUG_LOCKS_BACKTRACE is defined, which in turn is switched on when DEBUG=1.
(We disable DEBUG_LOCKS_BACKTRACE in some of the memory allocation tests because the locks used by the memory allocator take up too much room in the fake skiboot heap.)
libfdt: upgrade to upstream dtc.git 243176c
Upgrade libfdt/ to github.com/dgibson/dtc.git 243176c (“Fix bogus error on rebuild”)
This copies dtc/libfdt/ to skiboot/libfdt/, with the only change in that directory being the addition of README.skiboot and Makefile.inc.
This adds about 14kB text, 2.5kB compressed xz. This could be reduced or mostly eliminated by cutting out fdt version checks and unused code, but tracking upstream is a bigger benefit at the moment.
This loses commits:
14ed2b842f61 (“libfdt: add basic sanity check to fdt_open_into”)
bc7bb3d12bc1 (“sparse: fix declaration of fdt_strerror”)
As well as some prehistoric similar kinds of things, which is the punishment for us not being good downstream citizens and sending things upstream! Syncing to upstream will make that effort simpler in future.
General Fixes¶
Since skiboot v6.4-rc1:
libflash: Fix broken continuations
Some of the libflash debug messages don’t print a newlines at the end of the line and assume that the next print will be contigious with the last. This isn’t true in skiboot since log messages are prefixed with a timestamp. This results in funny looking output such as:
LIBFLASH: Verifying... LIBFLASH: reading page 0x01963000..0x01964000...[3.084846885,7] same ! LIBFLASH: reading page 0x01964000..0x01965000...[3.086164489,7] same !
Fix this by moving the “same !” debug message to a new line with the prefix “LIBFLASH: …” to indicate it’s a continuation of the last statement.
First reported in https://github.com/open-power/skiboot/issues/51