- 05 Aug, 2015 7 commits
-
-
Vitaly Kuznetsov authored
current_pt_regs() sometimes returns regs of the userspace process and in case of a kernel crash this is not what we need to report. E.g. when we trigger crash with sysrq we see the following: ... RIP: 0010:[<ffffffff815b8696>] [<ffffffff815b8696>] sysrq_handle_crash+0x16/0x20 RSP: 0018:ffff8800db0a7d88 EFLAGS: 00010246 RAX: 000000000000000f RBX: ffffffff820a0660 RCX: 0000000000000000 ... at the same time current_pt_regs() give us: ip=7f899ea7e9e0, ax=ffffffffffffffda, bx=26c81a0, cx=7f899ea7e9e0, ... These registers come from the userspace process triggered the crash. As we don't even know which process it was this information is rather useless. When kernel crash happens through 'die' proper regs are being passed to all receivers on the die_chain (and panic_notifier_list is being notified with the string passed to panic() only). If panic() is called manually (e.g. on BUG()) we won't get 'die' notification so keep the 'panic' notification reporter as well but guard against double reporting. Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
K. Y. Srinivasan <kys@microsoft.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Vitaly Kuznetsov authored
Full kernel hang is observed when kdump kernel starts after a crash. This hang happens in vmbus_negotiate_version() function on wait_for_completion() as Hyper-V host (Win2012R2 in my testing) never responds to CHANNELMSG_INITIATE_CONTACT as it thinks the connection is already established. We need to perform some mandatory minimalistic cleanup before we start new kernel. Reported-by:
-
Vitaly Kuznetsov authored
At the very late stage of kexec a driver (which are not being unloaded) can try to post a message or signal an event. This will crash the kernel as we already did hv_cleanup() and the hypercall page is NULL. Move all common (between 32 and 64 bit code) declarations to the beginning of the do_hypercall() function. Unfortunately we have to write the !hypercall_page check twice to not mix declarations and code. Signed-off-by:
-
Vitaly Kuznetsov authored
When general-purpose kexec (not kdump) is being performed in Hyper-V guest the newly booted kernel fails with an MCE error coming from the host. It is the same error which was fixed in the "Drivers: hv: vmbus: Implement the protocol for tearing down vmbus state" commit - monitor pages remain special and when they're being written to (as the new kernel doesn't know these pages are special) bad things happen. We need to perform some minimalistic cleanup before booting a new kernel on kexec. To do so we need to register a special machine_ops.shutdown handler to be executed before the native_machine_shutdown(). Registering a shutdown notification handler via the register_reboot_notifier() call is not sufficient as it happens to early for our purposes. machine_ops is not being exported to modules (and I don't think we want to export it) so let's do this in mshyperv.c The minimalistic cleanup consists of cleaning up clockevents, synic MSRs, guest os id MSR, and hypercall MSR. Kdump doesn't require all this stuff as it lives in a separate memory space. Signed-off-by:
-
Vitaly Kuznetsov authored
If some piece of code wants to check kexec_in_progress it has to be put in #ifdef CONFIG_KEXEC block to not break the build in !CONFIG_KEXEC case. Overcome this limitation by defining kexec_in_progress to false. Signed-off-by:
-
Vitaly Kuznetsov authored
We already have hv_synic_free() which frees all per-cpu pages for all CPUs, let's remove the hv_synic_free_cpu() call from hv_synic_cleanup() so it will be possible to do separate cleanup (writing to MSRs) and final freeing. This is going to be used to assist kexec. Signed-off-by:
-
Tomas Winkler authored
Remove bogus check on pm_runtime_active that prevented disconnection from a client in case the device was resuming from power gating but not yet active. Fix regression introduced by 18901357 mei: disconnect on connection request timeout Signed-off-by:
Tomas Winkler <tomas.winkler@intel.com> Signed-off-by:
Alexander Usyskin <alexander.usyskin@intel.com> Signed-off-by:
-
- 04 Aug, 2015 33 commits
-
-
Tomas Winkler authored
HBM 2.0 version for Sunrise point Skylake (PCH) based devices Signed-off-by:
-
Tomas Winkler authored
Add MEI devices ids for Intel Sunrisepoint Skylake (PCH) Signed-off-by:
-
Alexander Usyskin authored
A BIOS may put the device in d0i3 on platform initialization so it won’t consume power even if the driver is not present, in turn the driver has to wake up the devices on load in order to perform the initialization sequence and move it back to low power state on driver remove. Signed-off-by:
-
Alexander Usyskin authored
Move mei_me_hw_reset down in the source file to avoid forward declarations when introducing d0i3 flow in the next patch. Signed-off-by:
-
Alexander Usyskin authored
Rework mei power gating state machine to support entry and exit to and from D0i3 power state. The choice between legacy and D0i3 routines is conditioned on d0i3_supported flag. The patch introduces warning: drivers/misc/mei/hw-me.c:901:12: warning: ‘mei_me_d0i3_enter’ defined but not used [-Wunused-function] it will go away in consequent patch Signed-off-by:
-
Alexander Usyskin authored
Before adding support for D0i3 we need to reorganize the hbm pg handling Move HBM PG response code to dedicated functions in order to unclutter hbm command switch. Add check for the right system state before message processing and return -EPROTO in state mismatch case. Signed-off-by:
-
Alexander Usyskin authored
D0i3 adds additional interrupt reason bit, therefore we add a variable intr_source to save the interrupt causes for further dispatching. The interrupt cause is saved in the irq quick handler to achieve unified behavior for both MSI enabled and shared interrupt platforms. Signed-off-by:
-
Alexander Usyskin authored
Detect d0i3 low power state during hw configuration, the value is set in HFS_1 pci config reigister. Signed-off-by:
-
Tomas Winkler authored
Starting with Intel Sunrisepoint (Skylake PCH) the MEI device supports D0i3 low power state. Add D0i3 control registers. Signed-off-by:
-
Alexander Usyskin authored
The device can return error 5 (NOT_ALLOWED) on connection attempt. This error can happen if: 1. An another connection attempt is in progress 2. There is an attempt to connect a fixed (connectionless) client 3. The number of available connections is exceeded (new in HBM 2.0) We should not hit that error unless there is an internal book keeping hiccup except option (3), therefore we translate the error code to errno EBUSY; Signed-off-by:
-
Alexander Usyskin authored
Enable drivers on mei client bus to subscribe to asynchronous event notifications. Introduce events_mask to the existing callback infrastructure so it is possible to handle both RX and event notification. Signed-off-by:
-
Tomas Winkler authored
A process can be informed about client notification also via SIGIO with POLL_PRI event. Signed-off-by:
-
Tomas Winkler authored
Polling on priority events is translated on waiting for event notification. One need to enable notification prior for calling select or poll system call otherwise process will not wait. Signed-off-by:
-
Tomas Winkler authored
Add ioctl IOCTL_MEI_NOTIFY_SET for enabling and disabling async event notification. Add ioctl IOCTL_MEI_NOTIFY_GET for receiving and acking an event notification. Signed-off-by:
-
Tomas Winkler authored
mei_cl_notify_get is to be called by a host client to wait, receive, and ack the event notification. Signed-off-by:
-
Tomas Winkler authored
Add per client notification request infrastructure that allows client to enable or disable async event notification. Signed-off-by:
-
Tomas Winkler authored
Only FW version 2.0 and newer support the async event notification. For backward compatibility block the feature if the FW version is older then 2.0 Signed-off-by:
-
Tomas Winkler authored
Implement sending and reception handlers for the async event notification hbm commands. Add client notification book keeping data required for the messages notify_en to indicate whether notification is enabled notify_ev to indicate whether an event is pending Signed-off-by: -
Tomas Winkler authored
FW has gained new capability where a FW client can asynchronously notify the host that an event has occurred in its process. The notification doesn't provide any data and host may need to query further the FW client in order to get details of the event. Host can subscribe or unsubscribe to the event notification via designated HBM commands, and also the notification is carried on a new HBM command. This patch adds definitions of asynchronous notification HBM commands. Signed-off-by:
-
Alexander Usyskin authored
For the FW with HBM version >= 2.0 we don't need to reset the whole device in case of a particular client failing to connect, it is enough to send disconnect a request to bring the device to the stable state. Signed-off-by:
-
Tomas Winkler authored
HBM version 2.0 and above allows ME clients in the system to register/unregister after the system is fully initialized. Clients may be added or removed after enum_resp message was received 1. To preserve backward compatibility the driver can opt-in to receive client add messages by setting allow_add field in enum_req 2. A new client is added upon reception of MEI_HBM_ADD_CLIENT_REQ_CMD 3. A client is removed in a lazy manner when connection request respond with MEI_HBMS_CLIENT_NOT_FOUND status Signed-off-by:
-
Tomas Winkler authored
MEI bus was designed around nfc and was hard to extend. Instead of the hard coded way of adding the devices on the mei bus we scan the whole me client list and create a device for each eligible me client (mei_cl_bus_rescan); currently we support only clients with single connection and fixed address clients. NFC radio name detection is run as a fixup routine The patch replaces handling the device list based on struct me_cl to device list based on me_cl_devices. The creating a connection is pushed from the device creation time to device enablement. Signed-off-by:
-
Tomas Winkler authored
Remove the dependency on struct ndev from the nfc device name creation function so it is possible to use it in a fixup routine Signed-off-by:
-
Tomas Winkler authored
Currently we support only clients with single connection and fixed address clients so all other clients are blacklisted Signed-off-by:
-
Tomas Winkler authored
Blacklist nfc info client which is only used for retrieval of the NFC radio version Signed-off-by:
-
Tomas Winkler authored
Split the device registration into allocation and device struct initialization, device setup, and the final device registration. This why it is possible to run fixups and quirks during the setup stage on an initialized device. Each fixup routine effects do_match flag. If the flag is set to false at the end the device won't be registered on the bus. Signed-off-by:
-
Tomas Winkler authored
Instead of holding the list of host clients (me_cl) we want to keep the list me client devices (mei_cl_device) This way we can create host to me client connection only when needed. Add list head to mei_cl_device and cl_bus_lock Add bus_added flag to the me client (mei_me_client) to track if the appropriate mei_cl_device was already created and is_added flag to mei_cl_device to track if it was already added to the device list across the bus rescans Signed-off-by:
-
Tomas Winkler authored
Add reference to the bus device (mei_device) for easier access. To ensures that referencing cldev->bus is valid during cldev life time we increase the bus ref counter on a client device creation and drop it on the device release. Signed-off-by:
-
Tomas Winkler authored
Instead of generating device id on the fly during probing we find the matching id entry on the device id table. Get bus the module reference counter so it cannot be unloaded after the driver has bounded to the client device Signed-off-by:
-
Tomas Winkler authored
mei_cl_device_match now calls mei_cl_device_find that returns the matching device id in the device id table. We will utilize the mei_cl_device_find during probing to locate the matching entry. Signed-off-by:
-
Tomas Winkler authored
If event registeration has failed, the caller should know about it. Signed-off-by:
-
Tomas Winkler authored
Do not enable events implicitly in mei_cl_enable_device, it should be done explicitly using mei_cl_register_event_cb so the events are enabled only when needed. The NFC drivers has been already using it that way so no need for further changes just remove the code from mei_cl_enable_device. Signed-off-by:
-
Tomas Winkler authored
Rename mei_cl_uevent to mei_cl_device_uevent to match the naming convention of mei_cl_bus_type functions Signed-off-by:
-
