- 17 Jul, 2014 40 commits
-
-
Thomas Gleixner authored
commit c58d80f5 upstream. Some TI chips raise the DMA complete interrupt before the actual transfer has been completed. The code tries to busy wait for a few microseconds and if that fails it arms an hrtimer to recheck. So far so good, but that has the following issue: CPU 0 CPU1 start_next_transfer(RQ1); DMA interrupt if (premature_irq(RQ1)) if (!hrtimer_active(timer)) hrtimer_start(timer); hrtimer expires timer->state = CALLBACK_RUNNING; timer->fn() cppi41_recheck_tx_req() complete_request(RQ1); if (requests_pending()) start_next_transfer(RQ2); DMA interrupt if (premature_irq(RQ2)) if (!hrtimer_active(timer)) hrtimer_start(timer); timer->state = INACTIVE; The premature interrupt of request2 on CPU1 does not arm the timer and therefor the request completion never happens because it checks for !hrtimer_active(). hrtimer_active() evaluates: timer->state != HRTIMER_STATE_INACTIVE which of course evaluates to true in the above case as timer->state is CALLBACK_RUNNING. That's clearly documented: * A timer is active, when it is enqueued into the rbtree or the * callback function is running or it's in the state of being migrated * to another cpu. But that's not what the code wants to check. The code wants to check whether the timer is queued, i.e. whether its armed and waiting for expiry. We have a helper function for this: hrtimer_is_queued(). This evaluates: timer->state & HRTIMER_STATE_QUEUED So in the above case this evaluates to false and therefor forces the DMA interrupt on CPU1 to call hrtimer_start(). Use hrtimer_is_queued() instead of hrtimer_active() and evrything is good. Reported-by:
Torben Hohn <torbenh@linutronix.de> Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
Felipe Balbi <balbi@ti.com> Signed-off-by:
Jiri Slaby <jslaby@suse.cz>
-
Linus Walleij authored
commit 82363cf2 upstream. There is a regression in the upcoming v3.16-rc1, that is caused by a problem that has been around for a while but now finally hangs the system. The bootcrawl looks like this: pinctrl-nomadik soc:pinctrl: pin GPIO256_AF28 already requested by a03e0000.usb_per5; cannot claim for musb-hdrc.0.auto pinctrl-nomadik soc:pinctrl: pin-256 (musb-hdrc.0.auto) status -22 pinctrl-nomadik soc:pinctrl: could not request pin 256 (GPIO256_AF28) from group usb_a_1 on device pinctrl-nomadik musb-hdrc musb-hdrc.0.auto: Error applying setting, reverse things back HS USB OTG: no transceiver configured musb-hdrc musb-hdrc.0.auto: musb_init_controller failed with status -517 platform musb-hdrc.0.auto: Driver musb-hdrc requests probe deferral (...) The ux500 MUSB driver propagates the OF node to the dynamically created musb-hdrc device, which is incorrect as it makes the OF core believe there are two devices spun from the very same DT node, which confuses other parts of the device core, notably the pin control subsystem, which will try to apply all the pin control settings also to the HDRC device as it gets instantiated. (The OMAP2430 for example, does not set the of_node member.) Cc: Arnd Bergmann <arnd@arndb.de> Acked-by:
Lee Jones <lee.jones@linaro.org> Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Signed-off-by:
-
Bjørn Mork authored
commit b0ebef36 upstream. Adding a couple of Olivetti modems and blacklisting the net function on a couple which are already supported. Reported-by:
Lars Melin <larsm17@gmail.com> Signed-off-by:
Bjørn Mork <bjorn@mork.no> Signed-off-by:
Johan Hovold <johan@kernel.org> Signed-off-by:
-
Oliver Neukum authored
commit 1cab4c68 upstream. Reported by Alif Mubarak Ahmad: This device vendor and product id is 1c9e:9800 It is working as serial interface with generic usbserial driver. I thought it is more suitable to use usbserial option driver, which has better capability distinguishing between modem serial interface and micro sd storage interface. [ johan: style changes ] Signed-off-by:
Oliver Neukum <oneukum@suse.de> Tested-by:
Alif Mubarak Ahmad <alive4ever@live.com> Signed-off-by:
-
Wang, Yu authored
commit d6236f6d upstream. The system suspend flow as following: 1, Freeze all user processes and kenrel threads. 2, Try to suspend all devices. 2.1, If pci device is in RPM suspended state, then pci driver will try to resume it to RPM active state in the prepare stage. 2.2, xhci_resume function calls usb_hcd_resume_root_hub to queue two workqueue items to resume usb2&usb3 roothub devices. 2.3, Call suspend callbacks of devices. 2.3.1, All suspend callbacks of all hcd's children, including roothub devices are called. 2.3.2, Finally, hcd_pci_suspend callback is called. Due to workqueue threads were already frozen in step 1, the workqueue items can't be scheduled, and the roothub devices can't be resumed in this flow. The HCD_FLAG_WAKEUP_PENDING flag which is set in usb_hcd_resume_root_hub won't be cleared. Finally, hcd_pci_suspend will return -EBUSY, and system suspend fails. The reason why this issue doesn't show up very often is due to that choose_wakeup will be called in step 2.3.1. In step 2.3.1, if udev->do_remote_wakeup is not equal to device_may_wakeup(&udev->dev), then udev will resume to RPM active for changing the wakeup settings. This has been a lucky hit which hides this issue. For some special xHCI controllers which have no USB2 port, then roothub will not match hub driver due to probe failed. Then its do_remote_wakeup will be set to zero, and we won't be as lucky. xhci driver doesn't need to resume roothub devices everytime like in the above case. It's only needed when there are pending event TRBs. This patch should be back-ported to kernels as old as 3.2, that contains the commit f69e3120 "USB: XHCI: resume root hubs when the controller resumes" Signed-off-by:
Wang, Yu <yu.y.wang@intel.com> Acked-by:
Alan Stern <stern@rowland.harvard.edu> [use readl() instead of removed xhci_readl(), reword commit message -Mathias] Signed-off-by:
Mathias Nyman <mathias.nyman@linux.intel.com> Signed-off-by:
-
Mathias Nyman authored
commit 3213b151 upstream. The transfer burst count (TBC) field in xhci 1.0 hosts should be set to the number of bursts needed to transfer all packets in a isoc TD. Supported values are 0-2 (1 to 3 bursts per service interval). Formula for TBC calculation is given in xhci spec section 4.11.2.3: TBC = roundup( Transfer Descriptor Packet Count / Max Burst Size +1 ) - 1 This patch should be applied to stable kernels since 3.0 that contain the commit 5cd43e33 "xhci 1.0: Set transfer burst count field." Suggested-by:
ShiChun Ma <masc2008@qq.com> Signed-off-by:
-
Paolo Bonzini authored
commit 8faeb529 upstream. Even though the virtio-scsi spec guarantees that all requests related to the TMF will have been completed by the time the TMF itself completes, the request queue's callback might not have run yet. This causes requests to be completed more than once, and as a result triggers a variety of BUGs or oopses. Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com> Reviewed-by:
Venkatesh Srinivas <venkateshs@google.com> Signed-off-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
Ulrich Obergfell authored
commit 8922a908 upstream. After scsi_try_to_abort_cmd returns, the eh_abort_handler may have already found that the command has completed in the device, causing the host_byte to be nonzero (e.g. it could be DID_ABORT). When this happens, ORing DID_TIME_OUT into the host byte will corrupt the result field and initiate an unwanted command retry. Fix this by using set_host_byte instead, following the model of commit 2082ebc4. Signed-off-by:
Ulrich Obergfell <uobergfe@redhat.com> [Fix all instances according to review comments. - Paolo] Signed-off-by:
Ewan D. Milne <emilne@redhat.com> Reviewed-by:
Hannes Reinecke <hare@suse.de> Signed-off-by:
-
Paolo Bonzini authored
commit cdda0e5a upstream. Calling the workqueue interface on uninitialized work items isn't a good idea even if they're zeroed. It's not failing catastrophically only through happy accidents. Signed-off-by:
Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by:
-
Brian King authored
commit 7114aae0 upstream. Add a memory barrier prior to sending a new command to the VIOS to ensure the VIOS does not receive stale data in the command buffer. Also add a memory barrier when processing the CRQ for completed commands. Signed-off-by:
Brian King <brking@linux.vnet.ibm.com> Acked-by:
Nathan Fontenot <nfont@linux.vnet.ibm.com> Signed-off-by:
-
Brian King authored
commit 9ee75597 upstream. If a CRQ reset is triggered for some reason while in the middle of performing VSCSI adapter initialization, we don't want to call the done function for the initialization MAD commands as this will only result in two threads attempting initialization at the same time, resulting in failures. Signed-off-by:
-
Takashi Iwai authored
commit 92a586bd upstream. When a USB-audio device is disconnected while PCM is still running, we still see some race: the disconnect callback calls snd_usb_endpoint_free() that calls release_urbs() and then kfree() while a PCM stream would be closed at the same time and calls stop_endpoints() that leads to wait_clear_urbs(). That is, the EP object might be deallocated while a PCM stream is syncing with wait_clear_urbs() with the same EP. Basically calling multiple wait_clear_urbs() would work fine, also calling wait_clear_urbs() and release_urbs() would work, too, as wait_clear_urbs() just reads some fields in ep. The problem is the succeeding kfree() in snd_pcm_endpoint_free(). This patch moves out the EP deallocation into the later point, the destructor callback. At this stage, all PCMs must have been already closed, so it's safe to free the objects. Reported-by:
Takashi Iwai <tiwai@suse.de> Signed-off-by:
-
Oleg Nesterov authored
commit 4af4206b upstream. syscall_regfunc() and syscall_unregfunc() should set/clear TIF_SYSCALL_TRACEPOINT system-wide, but do_each_thread() can race with copy_process() and miss the new child which was not added to the process/thread lists yet. Change copy_process() to update the child's TIF_SYSCALL_TRACEPOINT under tasklist. Link: http://lkml.kernel.org/p/20140413185854.GB20668@redhat.com Fixes: a871bd33 "tracing: Add syscall tracepoints" Acked-by:
Frederic Weisbecker <fweisbec@gmail.com> Acked-by:
Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by:
Oleg Nesterov <oleg@redhat.com> Signed-off-by:
Steven Rostedt <rostedt@goodmis.org> Signed-off-by:
-
Steven Rostedt (Red Hat) authored
commit 379cfdac upstream. In order to prevent the saved cmdline cache from being filled when tracing is not active, the comms are only recorded after a trace event is recorded. The problem is, a comm can fail to be recorded if the trace_cmdline_lock is held. That lock is taken via a trylock to allow it to happen from any context (including NMI). If the lock fails to be taken, the comm is skipped. No big deal, as we will try again later. But! Because of the code that was added to only record after an event, we may not try again later as the recording is made as a oneshot per event per CPU. Only disable the recording of the comm if the comm is actually recorded. Fixes: 7ffbd48d "tracing: Cache comms only after an event occurred" Signed-off-by:
-
Jacob Keller authored
commit 8401aa1f upstream. Update the SubmittingPatches process to include howto about the new 'Fixes:' tag to be used when a patch fixes an issue in a previous commit (found by git-bisect for example). Signed-off-by:
Jacob Keller <jacob.e.keller@intel.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by:
-
Greg Kroah-Hartman authored
commit 4a3a9904 upstream. Jan points out that I forgot to make the needed fixes to the lz4_uncompress_unknownoutputsize() function to mirror the changes done in lz4_decompress() with regards to potential pointer overflows. The only in-kernel user of this function is the zram code, which only takes data from a valid compressed buffer that it made itself, so it's not a big issue. But due to external kernel modules using this function, it's better to be safe here. Reported-by:
Jan Beulich <JBeulich@suse.com> Cc: "Don A. Bailey" <donb@securitymouse.com> Signed-off-by:
-
Tejun Heo authored
commit b9cd18de upstream. The 'sysret' fastpath does not correctly restore even all regular registers, much less any segment registers or reflags values. That is very much part of why it's faster than 'iret'. Normally that isn't a problem, because the normal ptrace() interface catches the process using the signal handler infrastructure, which always returns with an iret. However, some paths can get caught using ptrace_event() instead of the signal path, and for those we need to make sure that we aren't going to return to user space using 'sysret'. Otherwise the modifications that may have been done to the register set by the tracer wouldn't necessarily take effect. Fix it by forcing IRET path by setting TIF_NOTIFY_RESUME from arch_ptrace_stop_needed() which is invoked from ptrace_stop(). Signed-off-by:
Tejun Heo <tj@kernel.org> Reported-by:
Andy Lutomirski <luto@amacapital.net> Acked-by:
-
Peter Christensen authored
commit f44a5f45 upstream. Receiving a ICMP response to an IPIP packet in a non-linear skb could cause a kernel panic in __skb_pull. The problem was introduced in commit f2edb9f7 ("ipvs: implement passive PMTUD for IPIP packets"). Signed-off-by:
Peter Christensen <pch@ordbogen.com> Acked-by:
Julian Anastasov <ja@ssi.bg> Signed-off-by:
Simon Horman <horms@verge.net.au> Signed-off-by:
-
Deng-Cheng Zhu authored
commit 8c9eb041 upstream. kvm_arch_vcpu_free() is called in 2 code paths: 1) kvm_vm_ioctl() kvm_vm_ioctl_create_vcpu() kvm_arch_vcpu_destroy() kvm_arch_vcpu_free() 2) kvm_put_kvm() kvm_destroy_vm() kvm_arch_destroy_vm() kvm_mips_free_vcpus() kvm_arch_vcpu_free() Neither of the paths handles VCPU free. We need to do it in kvm_arch_vcpu_free() corresponding to the memory allocation in kvm_arch_vcpu_create(). Signed-off-by:
Deng-Cheng Zhu <dengcheng.zhu@imgtec.com> Reviewed-by:
James Hogan <james.hogan@imgtec.com> Signed-off-by:
-
James Hogan authored
commit c6c0a663 upstream. The kfree() function already NULL checks the parameter so remove the redundant NULL checks before kfree() calls in arch/mips/kvm/. Signed-off-by:
-
Jeff Mahoney authored
commit 22e7478d upstream. Prior to commit 0e4f6a79 (Fix reiserfs_file_release()), reiserfs truncates serialized on i_mutex. They mostly still do, with the exception of reiserfs_file_release. That blocks out other writers via the tailpack mutex and the inode openers counter adjusted in reiserfs_file_open. However, NFS will call reiserfs_setattr without having called ->open, so we end up with a race when nfs is calling ->setattr while another process is releasing the file. Ultimately, it triggers the BUG_ON(inode->i_size != new_file_size) check in maybe_indirect_to_direct. The solution is to pull the lock into reiserfs_setattr to encompass the truncate_setsize call as well. Signed-off-by:
Jeff Mahoney <jeffm@suse.com> Signed-off-by:
Jan Kara <jack@suse.cz> Signed-off-by:
-
Scott Wood authored
commit 6663a4fa upstream. Commit 59a53afe "powerpc: Don't setup CPUs with bad status" broke ePAPR SMP booting. ePAPR says that CPUs that aren't presently running shall have status of disabled, with enable-method being used to determine whether the CPU can be enabled. Fix by checking for spin-table, which is currently the only supported enable-method. Signed-off-by:
Scott Wood <scottwood@freescale.com> Cc: Michael Neuling <mikey@neuling.org> Cc: Emil Medve <Emilian.Medve@Freescale.com> Signed-off-by:
Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by:
-
Benjamin Herrenschmidt authored
commit dd58a092 upstream. The Vector Crypto category instructions are supported by current POWER8 chips, advertise them to userspace using a specific bit to properly differentiate with chips of the same architecture level that might not have them. Signed-off-by:
-
Michael Neuling authored
commit 59a53afe upstream. OPAL will mark a CPU that is guarded as "bad" in the status property of the CPU node. Unfortunatley Linux doesn't check this property and will put the bad CPU in the present map. This has caused hangs on booting when we try to unsplit the core. This patch checks the CPU is avaliable via this status property before putting it in the present map. Signed-off-by:
Michael Neuling <mikey@neuling.org> Tested-by:
Anton Blanchard <anton@samba.org> Signed-off-by:
-
Paul Bolle authored
commit b69a1da9 upstream. Commit cd64d169 ("powerpc: mtmsrd not defined") added a check for CONFIG_PPC_CPU were a check for CONFIG_PPC_FPU was clearly intended. Fixes: cd64d169 ("powerpc: mtmsrd not defined") Signed-off-by:
Paul Bolle <pebolle@tiscali.nl> Signed-off-by:
-
Michael Ellerman authored
commit 3df48c98 upstream. In commit 330a1eb7 "Core EBB support for 64-bit book3s" I messed up clear_task_ebb(). It clears some but not all of the task's Event Based Branch (EBB) registers when we duplicate a task struct. That allows a child task to observe the EBBHR & EBBRR of its parent, which it should not be able to do. Fix it by clearing EBBHR & EBBRR. Signed-off-by:
Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
-
Paul Bolle authored
commit 6e0fdf9a upstream. Commit b0d278b7 ("powerpc/perf_event: Reduce latency of calling perf_event_do_pending") added a check for CONFIG_PMAC were a check for CONFIG_PPC_PMAC was clearly intended. Fixes: b0d278b7 ("powerpc/perf_event: Reduce latency of calling perf_event_do_pending") Signed-off-by:
-
Anton Blanchard authored
commit 5d73320a upstream. commit 8f9c0119 (compat: fs: Generic compat_sys_sendfile implementation) changed the PowerPC 64bit sendfile call from sys_sendile64 to sys_sendfile. Unfortunately this broke sendfile of lengths greater than 2G because sys_sendfile caps at MAX_NON_LFS. Restore what we had previously which fixes the bug. Signed-off-by:
-
Benjamin Herrenschmidt authored
commit c4cad90f upstream. We had a mix & match of flags used when creating legacy ports depending on where we found them in the device-tree. Among others we were missing UPF_SKIP_TEST for some kind of ISA ports which is a problem as quite a few UARTs out there don't support the loopback test (such as a lot of BMCs). Let's pick the set of flags used by the SoC code and generalize it which means autoconf, no loopback test, irq maybe shared and fixed port. Sending to stable as the lack of UPF_SKIP_TEST is breaking serial on some machines so I want this back into distros Signed-off-by:
-
Michael Ellerman authored
commit 09567e7f upstream. We have a bug in our hugepage handling which exhibits as an infinite loop of hash faults. If the fault is being taken in the kernel it will typically trigger the softlockup detector, or the RCU stall detector. The bug is as follows: 1. mmap(0xa0000000, ..., MAP_FIXED | MAP_HUGE_TLB | MAP_ANONYMOUS ..) 2. Slice code converts the slice psize to 16M. 3. The code on lines 539-540 of slice.c in slice_get_unmapped_area() synchronises the mm->context with the paca->context. So the paca slice mask is updated to include the 16M slice. 3. Either: * mmap() fails because there are no huge pages available. * mmap() succeeds and the mapping is then munmapped. In both cases the slice psize remains at 16M in both the paca & mm. 4. mmap(0xa0000000, ..., MAP_FIXED | MAP_ANONYMOUS ..) 5. The slice psize is converted back to 64K. Because of the check on line 539 of slice.c we DO NOT update the paca->context. The paca slice mask is now out of sync with the mm slice mask. 6. User/kernel accesses 0xa0000000. 7. The SLB miss handler slb_allocate_realmode() **uses the paca slice mask** to create an SLB entry and inserts it in the SLB. 18. With the 16M SLB entry in place the hardware does a hash lookup, no entry is found so a data access exception is generated. 19. The data access handler calls do_page_fault() -> handle_mm_fault(). 10. __handle_mm_fault() creates a THP mapping with do_huge_pmd_anonymous_page(). 11. The hardware retries the access, there is still nothing in the hash table so once again a data access exception is generated. 12. hash_page() calls into __hash_page_thp() and inserts a mapping in the hash. Although the THP mapping maps 16M the hashing is done using 64K as the segment page size. 13. hash_page() returns immediately after calling __hash_page_thp(), skipping over the code at line 1125. Resulting in the mismatch between the paca->context and mm->context not being detected. 14. The hardware retries the access, the hash it generates using the 16M SLB entry does NOT match the hash we inserted. 15. We take another data access and go into __hash_page_thp(). 16. We see a valid entry in the hpte_slot_array and so we call updatepp() which succeeds. 17. Goto 14. We could fix this in two ways. The first would be to remove or modify the check on line 539 of slice.c. The second option is to cause the check of paca psize in hash_page() on line 1125 to also be done for THP pages. We prefer the latter, because the check & update of the paca psize is not done until we know it's necessary. It's also done only on the current cpu, so we don't need to IPI all other cpus. Without further rearranging the code, the simplest fix is to pull out the code that checks paca psize and call it in two places. Firstly for THP/hugetlb, and secondly for other mappings as before. Thanks to Dave Jones for trinity, which originally found this bug. Signed-off-by:
Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Signed-off-by:
-
Gavin Shan authored
commit 54f112a3 upstream. In pseries_eeh_get_state(), EEH_STATE_UNAVAILABLE is always overwritten by EEH_STATE_NOT_SUPPORT because of the missed "break" there. The patch fixes the issue. Reported-by:
Joe Perches <joe@perches.com> Signed-off-by:
Gavin Shan <gwshan@linux.vnet.ibm.com> Signed-off-by:
-
Scott Mayhew authored
commit 18dd78c4 upstream. NFS_INO_INVALID_DATA cannot be ignored, even if we have a delegation. We're still having some problems with data corruption when multiple clients are appending to a file and those clients are being granted write delegations on open. To reproduce: Client A: vi /mnt/`hostname -s` while :; do echo "XXXXXXXXXXXXXXX" >>/mnt/file; sleep $(( $RANDOM % 5 )); done Client B: vi /mnt/`hostname -s` while :; do echo "YYYYYYYYYYYYYYY" >>/mnt/file; sleep $(( $RANDOM % 5 )); done What's happening is that in nfs_update_inode() we're recognizing that the file size has changed and we're setting NFS_INO_INVALID_DATA accordingly, but then we ignore the cache_validity flags in nfs_write_pageuptodate() because we have a delegation. As a result, in nfs_updatepage() we're extending the write to cover the full page even though we've not read in the data to begin with. Signed-off-by:
Scott Mayhew <smayhew@redhat.com> Signed-off-by:
Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by:
-
Mateusz Guzik authored
commit a914722f upstream. Otherwise the kernel oopses when remounting with IPv6 server because net is dereferenced in dev_get_by_name. Use net ns of current thread so that dev_get_by_name does not operate on foreign ns. Changing the address is prohibited anyway so this should not affect anything. Signed-off-by:
Mateusz Guzik <mguzik@redhat.com> Cc: linux-nfs@vger.kernel.org Cc: linux-kernel@vger.kernel.org Signed-off-by:
-
Trond Myklebust authored
commit 43b6535e upstream. Fix a bug, whereby nfs_update_inode() was declaring the inode to be up to date despite not having checked all the attributes. The bug occurs because the temporary variable in which we cache the validity information is 'sanitised' before reapplying to nfsi->cache_validity. Reported-by:
Kinglong Mee <kinglongmee@gmail.com> Signed-off-by:
-
Christoph Hellwig authored
commit 12337901 upstream. Note nobody's ever noticed because the typical client probably never requests FILES_AVAIL without also requesting something else on the list. Signed-off-by:
J. Bruce Fields <bfields@redhat.com> Signed-off-by:
-
Trond Myklebust authored
commit 6df200f5 upstream. Return the NULL pointer when the allocation fails. Reported-by:
Fengguang Wu <fengguang.wu@intel.com> Signed-off-by:
-
Trond Myklebust authored
commit c789102c upstream. If the accept() call fails, we need to put the module reference. Signed-off-by:
-
Bart Van Assche authored
commit 60e1751c upstream. Avoid that closing /dev/infiniband/umad<n> or /dev/infiniband/issm<n> triggers a use-after-free. __fput() invokes f_op->release() before it invokes cdev_put(). Make sure that the ib_umad_device structure is freed by the cdev_put() call instead of f_op->release(). This avoids that changing the port mode from IB into Ethernet and back to IB followed by restarting opensmd triggers the following kernel oops: general protection fault: 0000 [#1] PREEMPT SMP RIP: 0010:[<ffffffff810cc65c>] [<ffffffff810cc65c>] module_put+0x2c/0x170 Call Trace: [<ffffffff81190f20>] cdev_put+0x20/0x30 [<ffffffff8118e2ce>] __fput+0x1ae/0x1f0 [<ffffffff8118e35e>] ____fput+0xe/0x10 [<ffffffff810723bc>] task_work_run+0xac/0xe0 [<ffffffff81002a9f>] do_notify_resume+0x9f/0xc0 [<ffffffff814b8398>] int_signal+0x12/0x17 Reference: https://bugzilla.kernel.org/show_bug.cgi?id=75051Signed-off-by:
Bart Van Assche <bvanassche@acm.org> Reviewed-by:
Yann Droneaud <ydroneaud@opteya.com> Signed-off-by:
Roland Dreier <roland@purestorage.com> Signed-off-by:
-
Bart Van Assche authored
commit 8ec0a0e6 upstream. Avoid leaking a kref count in ib_umad_open() if port->ib_dev == NULL or if nonseekable_open() fails. Avoid leaking a kref count, that sm_sem is kept down and also that the IB_PORT_SM capability mask is not cleared in ib_umad_sm_open() if nonseekable_open() fails. Since container_of() never returns NULL, remove the code that tests whether container_of() returns NULL. Moving the kref_get() call from the start of ib_umad_*open() to the end is safe since it is the responsibility of the caller of these functions to ensure that the cdev pointer remains valid until at least when these functions return. Signed-off-by:
-
Bart Van Assche authored
commit 024ca901 upstream. Avoid that the loops that iterate over the request ring can encounter a pointer to a SCSI command in req->scmnd that is no longer associated with that request. If the function srp_unmap_data() is invoked twice for a SCSI command that is not in flight then that would cause ib_fmr_pool_unmap() to be invoked with an invalid pointer as argument, resulting in a kernel oops. Reported-by:
Sagi Grimberg <sagig@mellanox.com> Reference: http://thread.gmane.org/gmane.linux.drivers.rdma/19068/focus=19069Signed-off-by:
-
