- 24 Aug, 2016 29 commits
-
-
Jitendra Bhivare authored
In very rare scenario, connection gets killed after throwing this error: scsi host0: BM_2312 : Event CXN_KILLED_BAD_WRB_INDEX_ERROR[15]... CID : 4 connection1:0: detected conn error (1011) memset ISCSI_WRB descriptor to zero for all allocations of WRB handle. Signed-off-by:
Jitendra Bhivare <jitendra.bhivare@broadcom.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
Jitendra Bhivare authored
BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by:
Hannes Reinecke <hare@suse.com> Signed-off-by:
-
Jitendra Bhivare authored
Driver unload should call COMMON_FUNCTION_RESET. For TPE feature, this ensures that FW has knowledge about driver getting unloaded and can reset its bit vector. Signed-off-by:
-
Jitendra Bhivare authored
Sessions are no longer valid, so schedule sess_work to fail the sessions immediately when error is detected. This is done to avoid iSCSI transport layer to keep sending NOP-Out which driver any ways fail. Schedule sess_work immediately in case of HBA error. Old sessions are gone for good and need to be re-established. iscsi_session_failure needs process context hence this work. Signed-off-by:
-
Jitendra Bhivare authored
After UE is detected, check for recoverable error by reading SLIPORT SEMAPHORE register. If transient parity error i.e. 0xExxx then schedule recovery work on driver wq. FLag this error to prevent any transactions for the duration of ue2rp to restart polling. After that, if FW becomes ready then recover port. Wake up processes in wq before going offline. Wait for process to execute before cleaning up. Signed-off-by:
-
Jitendra Bhivare authored
mgmt_epfw_cleanup does not implement v1 of OPCODE_COMMON_ISCSI_CLEANUP IOCTL for SkyHawk. Replace use of MCCQ with BMBX for issuing the IOCTL. Remove be_mcc_compl_poll which is no longer needed. Signed-off-by:
-
Jitendra Bhivare authored
SLIPORT FUNCTION_RESET does not reset the chip. So POST status needs to be checked before issuing FUNCTION_RESET. The completion of FUNCTION_RESET is indicated in BMBX Rdy bit. be_cmd_fw_initialize too needs to be done before issuing any cmd to FW. be_cmd_fw_initialize is renamed as beiscsi_cmd_special_wrb. Rearrange and rename few functions in init and cleanup path. Signed-off-by:
-
Jitendra Bhivare authored
beiscsi_fail_session is defined in be_cmds.c: move it to be_iscsi.c Move card configuration commands to be_cmds.c. Signed-off-by:
-
Jitendra Bhivare authored
BE3 and SH cards can recover from transient parity errors treated earlier as unrecoverable errors. Add IOCTL to query FW support for this feature. Signed-off-by:
-
Jitendra Bhivare authored
UE detection in health check is done in a work scheduled in global wq. UE caused due to transient parity errors are recoverable and reported within 1s. If this check for TPE gets delayed, PF0 might initiate soft-reset and then status of UE recoverable is lost. Handle UE detection in timer routine. Move out EQ delay update work from health check. Make the IOCTL for EQ delay update non-blocking as the completion status is ignored. Signed-off-by:
-
Jitendra Bhivare authored
Boot work involves: 1. Find and fetch configured boot session and its handle. 2. Attempt to open the session if its not. 3. Get the session details for boot kset creation. 4. Logout of that session owned by FW. 5. Create boot kset for session details. All these actions were done in blocking call with retries in global wq. Other works in wq suffered if the IOCTLs stalled or timed out. This change moves all the boot work to make it non-blocking. The work queued in global wq just issues the IOCTL depending on the action to be taken and mcc wq schedules work depending on status of the IOCTL. Initial boot_work is started on link and ASYNC event. The other code changes move all boot related functions in one place and follow naming conventions. Signed-off-by:
-
Jitendra Bhivare authored
Save ue_detected and fw_timeout errors in state field of beiscsi_hba. BEISCSI_HBA_RUNNING BEISCSI_HBA_LINK_UP BEISCSI_HBA_BOOT_FOUND BEISCSI_HBA_PCI_ERR BEISCSI_HBA_FW_TIMEOUT BEISCSI_HBA_IN_UE Make sure no PCI transaction happens once in error state. Add checks in IO path to detect HBA in error. Skip hwi_purge_eq step which can't be done in error state. Signed-off-by:
-
Jitendra Bhivare authored
todo_mcc_cq is not needed as only MCC work is queued. todo_cq is not used at all. Rename functions to be consistent. Signed-off-by:
-
Jitendra Bhivare authored
alloc_mcc_tag was replaced with alloc_mcc_wrb and is no more used. beiscsi_pci_soft_reset is not used at all and won't be needed. Signed-off-by:
-
Jitendra Bhivare authored
Redefine FW IP types. Before issuing IOCTL to clear IP, check if IP is all zeroes. All zeroes IP implies IP is not set in FW so FW fails that IOCTL. Signed-off-by:
-
Jitendra Bhivare authored
Wrong settings displayed for iface: iface.header_digest = 192.168.197.22 iface.data_digest = 255.255.255.0 iface.immediate_data = 192.168.197.1 Process ISCSI_NET_PARAM only in beiscsi_iface_get_param. Signed-off-by:
-
Jitendra Bhivare authored
Rename mgmt_get_if_info to be consistent with APIs name. Rename create/destroy APIs to indicate IFACE operations. Remove legacy be2iscsi and use beiscsi. Signed-off-by:
-
Jitendra Bhivare authored
Move mgmt_get_all_if_id before any set param operation. Rename mgmt_get_all_if_id to beiscsi_if_get_handle. Signed-off-by:
-
Jitendra Bhivare authored
VLAN tag is L2 construct, move VLAN code out from configuring IP. Rearrange and rename the APIs to make it consistent. Replace ENOSYS with EPERM. Signed-off-by:
-
Jitendra Bhivare authored
If BOOTPROTO is changed to static, the DHCP IP address should be released. All cases are being handled mgmt_set_ip and mgmt_static_ip_modify. Rearrange IFACE APIs to: beiscsi_if_clr_ip beiscsi_if_set_ip beiscsi_if_en_static beiscsi_if_en_dhcp This simplifies release of DHCP IP when BOOTPROTO is set to static. Signed-off-by:
-
Jitendra Bhivare authored
Gateway APIs assume IP type as IPv4. Modify it to be generic to allow clearing of IPv6 gateway set using BIOS. Signed-off-by:
-
Jitendra Bhivare authored
ipv4_iface and ipv6_iface fields need to be set to NULL when destroyed. Before creation these are checked. Use these to report correct states. Signed-off-by:
-
Jitendra Bhivare authored
Driver takes significant time to load 1m:20s and unload 40s. Checkpatch script threw warning: WARNING: msleep < 20ms can sleep for up to 20ms; see Documentation/timers/timers-howto.txt To eliminate this warning msleep(1) was replaced with msleep(20) before submitting. msleep(20) in init and uninit path for creation and destroying of number of WRBQs, CQs, and EQs is adding to load/unload time. Replace msleep with schedule_timeout_uninterruptible of 1ms as its enough in most cases. Signed-off-by:
-
Jitendra Bhivare authored
This got unnecessarily introduced with other changes in previous commits. mcc_lock is taken only in process contexts. Signed-off-by:
-
Jitendra Bhivare authored
Following configuration is created with what driver exports: iface.vlan_id = 65535 iface.vlan_priority = 255 iface.vlan_state = <empty> vlan_state is empty as iscsiadm doesn't process "Disabled". When applying this configuration, iscsiadm checks for if vlan_state is "disable" if not it enables with value in vlan_id. 65535 not being valid value, 0 is applied. Use "enable" or "disable" for ISCSI_NET_PARAM. Signed-off-by:
-
Johannes Thumshirn authored
Provide a translation table between Ethernet and FC port speeds so odd speeds (from a Ethernet POV) like 8 Gbit are correctly mapped to sysfs and open-fcoe's fcoeadm. Before: Description: BCM57840 NetXtreme II 10/20-Gigabit Ethernet Revision: 11 Manufacturer: Broadcom Corporation Serial Number: 6CC2173EA1D0 Driver: bnx2x 1.712.30-0 Number of Ports: 1 Symbolic Name: bnx2fc (QLogic BCM57840) v2.10.3 over eth2 OS Device Name: host1 Node Name: 0x20006cc2173ea1d1 Port Name: 0x10006cc2173ea1d1 FabricName: 0x100000c0dd0ce717 Speed: unknown Supported Speed: 1 Gbit, 10 Gbit MaxFrameSize: 2048 bytes FC-ID (Port ID): 0x660702 State: Online After: Description: BCM57840 NetXtreme II 10/20-Gigabit Ethernet Revision: 11 Manufacturer: Broadcom Corporation Serial Number: 6CC2173EA1D0 Driver: bnx2x 1.712.30-0 Number of Ports: 1 Symbolic Name: bnx2fc (QLogic BCM57840) v2.10.3 over eth2 OS Device Name: host1 Node Name: 0x20006cc2173ea1d1 Port Name: 0x10006cc2173ea1d1 FabricName: 0x100000c0dd0ce717 Speed: 8 Gbit Supported Speed: 1 Gbit, 10 Gbit MaxFrameSize: 2048 bytes FC-ID (Port ID): 0x660701 State: Online Signed-off-by:Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by:
Hannes Reinicke <hare@suse.de> Reviewed-by:
Lee Duncan <lduncan@suse.com> Signed-off-by:
-
Matthew R. Ochs authored
Update the block library link in the API documentation. Signed-off-by:
Matthew R. Ochs <mrochs@linux.vnet.ibm.com> Acked-by:
Manoj N. Kumar <manoj@linux.vnet.ibm.com> Signed-off-by:
-
Matthew R. Ochs authored
The adapter file descriptor was previously cached within the kernel for a given context in order to support performing a close on behalf of an application. This is no longer needed as applications are now required to perform a close on the adapter file descriptor. Inspired-by:
Al Viro <viro@zeniv.linux.org.uk> Signed-off-by:
-
Matthew R. Ochs authored
Caching the adapter file descriptor and performing a close on behalf of an application is a poor design. This is due to the fact that once a file descriptor in installed, it is free to be altered without the knowledge of the cxlflash driver. This can lead to inconsistencies between the application and kernel. Furthermore, the nature of the former design is more exploitable and thus should be abandoned. To support applications performing a close on the adapter file that is associated with a context, a new flag is introduced to the user API to indicate to applications that they are responsible for the close following the cleanup (detach) of a context. The documentation is also updated to reflect this change in behavior. Inspired-by:
-
- 19 Aug, 2016 7 commits
-
-
Matthew R. Ochs authored
Currently, context user references are tracked via the list of LUNs that have attached to the context. While convenient, this is not intuitive without a deep study of the code and is inconsistent with the existing reference tracking patterns within the kernel. This design choice can lead to future bug injection. To improve code comprehension and better protect against future bugs, add explicit reference counting to contexts and migrate the context removal code to the kref release handler. Inspired-by:
-
Matthew R. Ochs authored
The context removal routine requires access to the owning adapter structure to reset the context within the AFU as part of the tear down sequence. In order to support kref adoption, the owning adapter must be accessible from the release handler. As the kref framework only provides the kref reference as the sole parameter, another means is needed to derive the owning adapter. As a remedy, the owning adapter reference is saved off within the context during initialization. Signed-off-by:
-
Matthew R. Ochs authored
Context information structures are protected by a mutex that is held when accessing/manipulating the context. When the code that manages these structures was authored, a decision was made to include taking the mutex as part of the allocation/initialization sequence and also handle the scenario where the mutex was already held when freeing the context. While not a problem outright, this design decision has been deemed as too flexible and the code should be made more rigid to avoid future bugs. In addition, further review of the code yields that the existing mutex manipulations in both of these context management paths are superfluous. This commit removes the obtaining of the context mutex in the context initialization routine and assumes the mutex is not held in the context free path. Inspired-by:
-
Hannes Reinecke authored
When all exchanges are reset the upper layers have already logged out of the remote port, so the exchanges can be reset without sending any ABTS. Signed-off-by:
Chad Dupuis <chad.dupuis@qlogic.com> Tested-by:
-
Hannes Reinecke authored
FC-LS mandates that we should invalidate all sequences before sending a LOGO. And we should set the event to RPORT_EV_STOP when a LOGO request has been received to signal that all exchanges are terminated. Signed-off-by:
-
Hannes Reinecke authored
When running in point-to-multipoint mode PLOGI is done after FLOGI completed. So when the PLOGI fails we should be sending a LOGO to the remote port. [mkp: Applied by hand] Signed-off-by:
-
Hannes Reinecke authored
When receiving a PRLO it just means that the operating parameters have changed, it does _not_ mean that the port doesn't want to communicate with us. So instead of implicitly logging out we should be issueing a PRLI to figure out the new operating parameters. We can always recover once PRLI fails. Signed-off-by:
-
- 12 Aug, 2016 4 commits
-
-
Steffen Maier authored
This was lost with commit 2c55b750 ("[SCSI] zfcp: Redesign of the debug tracing for SAN records.") but is necessary for problem determination, e.g. to see the currently active zone set during automatic port scan. For the large GPN_FT response (4 pages), save space by not dumping any empty residual entries. Signed-off-by:
Steffen Maier <maier@linux.vnet.ibm.com> Fixes: 2c55b750 ("[SCSI] zfcp: Redesign of the debug tracing for SAN records.") Cc: <stable@vger.kernel.org> #2.6.38+ Reviewed-by:
Alexey Ishchuk <aishchuk@linux.vnet.ibm.com> Reviewed-by:
Benjamin Block <bblock@linux.vnet.ibm.com> Reviewed-by:
-
Steffen Maier authored
commit 2c55b750 ("[SCSI] zfcp: Redesign of the debug tracing for SAN records.") started to add FC_CT_HDR_LEN which made zfcp dump random data out of bounds for RSPN GS responses because u.rspn.rsp is the largest and last field in the union of struct zfcp_fc_req. Other request/response types only happened to stay within bounds due to the padding of the union or due to the trace capping of u.gspn.rsp to ZFCP_DBF_SAN_MAX_PAYLOAD. Timestamp : ... Area : SAN Subarea : 00 Level : 1 Exception : - CPU id : .. Caller : ... Record id : 2 Tag : fsscth2 Request id : 0x... Destination ID : 0x00fffffc Payload short : 01000000 fc020000 80020000 00000000 xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx <=== 00000000 00000000 00000000 00000000 Payload length : 32 <=== struct zfcp_fc_req { [0] struct zfcp_fsf_ct_els ct_els; [56] struct scatterlist sg_req; [96] struct scatterlist sg_rsp; union { struct {req; rsp;} adisc; SIZE: 28+28= 56 struct {req; rsp;} gid_pn; SIZE: 24+20= 44 struct {rspsg; req;} gpn_ft; SIZE: 40*4+20=180 struct {req; rsp;} gspn; SIZE: 20+273= 293 struct {req; rsp;} rspn; SIZE: 277+16= 293 [136] } u; } SIZE: 432 Signed-off-by:
-
Steffen Maier authored
With commit 2c55b750 ("[SCSI] zfcp: Redesign of the debug tracing for SAN records.") we lost the N_Port-ID where an ELS response comes from. With commit 7c7dc196 ("[SCSI] zfcp: Simplify handling of ct and els requests") we lost the N_Port-ID where a CT response comes from. It's especially useful if the request SAN trace record with D_ID was already lost due to trace buffer wrap. GS uses an open WKA port handle and ELS just a D_ID, and only for ELS we could get D_ID from QTCB bottom via zfcp_fsf_req. To cover both cases, add a new field to zfcp_fsf_ct_els and fill it in on request to use in SAN response trace. Strictly speaking the D_ID on SAN response is the FC frame's S_ID. We don't need a field for the other end which is always us. Signed-off-by:
-
Steffen Maier authored
This information was lost with commit a54ca0f6 ("[SCSI] zfcp: Redesign of the debug tracing for HBA records.") but is required to debug e.g. invalid handle situations. Signed-off-by:
-
