Technical Notes. BZ#8. Suspending a system (mode S3) running on a HP Z1 All- in- one Workstation with an internal Embedded Display. Port (e. DP) panel and an external Display. Port (DP) monitor, and, consequently, waking up the system caused the backlight of the e. DP panel to not be re- enabled. To fix this issue, the code that handles suspending in the i. BLC_PWM_CPU_CTL parameter using the I9. WRITE function after writing the BLC_PWM_CPU_CTL2 parameter. It affects all versions of Firebird up to and including v.2.0.5 and v.2.1.3. Editions of the v.2.1 release notes are merged with the notes. Support for 64-bit. Manok mo sagot ko hardening - Search files on torrent trackers without registration and rating. Ko-approach-64-bit-0 5-build-284. 904 KB. 302 How can I detect in the.pro file if I am compiling on a 32 bit or a 64 bit platform? 306 How can I build 64 bit Qt on a 32 bit., 10.0, 5.0); painter. Ko Approach (64-bit) 0.5 Build 284 International TractorPrior to this update, when a VLAN device was set up on a qlge interface, running the TCP Stream Performance test using the netperf utility to test TCP/IPv. This was due to an unsupported feature (NETIF_F_IPV6_CSUM) which was enabled via the NETIF_F_TSO6 flag. This update removes the NETIF_F_TSO6 flag from qlge code and TCP/IPv. The isci driver copied the result of a "Register Device to Host" frame into the wrong buffer causing the SATA DOWNLOAD MICROCODE command to fail, preventing the download of hard drive firmware. This bug in the frame handler routine caused a timeout, resulting in a reset. With this update, the underlying source code has been modified to address this issue, and the isci driver successfully completes SATA DOWNLOAD MICROCODE commands as expected. In the x. HCI code, due to a descriptor that incorrectly pointed at the USB 3. USB 2. 0 registers, kernel panic could occur when more USB 2. USB 3. 0 registers. This update fixes the descriptor to point at the USB 2. When the "perf script - -gen- script" command was called with a perf. NULL "pevent" pointer. With this update, the underlying source code has been modified to address this issue, and the aforementioned command no longer crashes. Running the mq_notify/5- 1 test case from the Open POSIX test suite resulted in corrupted memory, later followed by various kernel crash/BUG messages. This update addresses the mq_send/receive memory corruption issue in the inter- process communication code, and the aforementioned test case no longer fails. Bond masters and slaves now have separate VLAN groups. As such, if a slave device incurred a network event that resulted in a failover, the VLAN device could process this event erroneously. With this update, when a VLAN is attached to a master device, it ignores events generated by slave devices so that the VLANs do not go down until the bond master does. Previously in the kernel, when the leap second hrtimer was started, it was possible that the kernel livelocked on the xtime_lock variable. This update fixes the problem by using a mixture of separate subsystem locks (timekeeping and ntp) and removing the xtime_lock variable, thus avoiding the livelock scenarios that could occur in the kernel. After the leap second was inserted, applications calling system calls that used futexes consumed almost 1. CPU time. This occurred because the kernel's timekeeping structure update did not properly update these futexes. The futexes repeatedly expired, re- armed, and then expired immediately again. This update fixes the problem by properly updating the futex expiration times by calling the clock_was_set_delayed() function, an interrupt- safe method of the clock_was_set() function. When the Fibre Channel (FC) layer sets a device to "running", the layer also scans for other new devices. Previously, there was a race condition between these two operations. Consequently, for certain targets, thousands of invalid devices were created by the SCSI layer and the udev service. This update ensures that the FC layer always sets a device to "online" before scanning for others, thus fixing this bug. Certain disk device arrays report a medium error without returning any data. This was not being handled correctly in cases where low level device drivers were not setting the optional residual field, however, most modern low level drivers do set it. This update correctly handles cases where low level drivers do not set the residual field in the upper level sd driver, avoiding the potential data corruption. This update reverts a previously- applied patch that caused the qla. IBM POWER 7 7. 89. X system. This patch has also been isolated as the cause of Dynamic Logical Partitioning (DLPAR) memory remove failures on 2 adapters. Previous update changed the /proc/stat code to use the get_cpu_idle_time_us() and get_cpu_iowait_time_us() macros if dynamic ticks are enabled in the kernel. This could lead to problems on IBM System z architecture that defines the arch_idle_time() macro. For example, executing the "vmstat" command could fail with "Floating point exception" followed by a core dump. The underlying source code has been modified so that the arch_idle_time() macro is used for idle and iowait times, which prevents the mentioned problem. The Stream Control Transmission Protocol (SCTP) process became unresponsive inside the sctp_wait_for_sndbuf() function when the sender exhausted the send buffer and waited indefinitely to be woken up. This was because twice the amount of data was accounted for during a packet transmission, once when constructing the packet and the second time when transmitting it. Thus, the available memory resources were used up too early, causing a deadlock. With this update, only a single byte is reserved to ensure the socket stays alive for the life time of the packet, and the SCTP process no longer hangs. If there are no active threads using a semaphore, blocked threads should be unblocked. Previously, the R/W semaphore code looked for a semaphore counter as a whole to reach zero - which is incorrect because at least one thread is usually queued on the semaphore and the counter is marked to reflect this. As a consequence, the system could become unresponsive when an application used direct I/O on the XFS file system. With this update, only the count of active semaphores is checked, thus preventing the hang in this scenario. When performing PCI device assignment on AMD systems, a virtual machine using the assigned device could not be able to boot, as the device had failed the assignment, leaving the device in an unusable state. This was due to an improper range check that omitted the last PCI device in a PCI subsystem or tree. The check has been fixed to include the full range of PCI devices in a PCI subsystem or tree. This bug fix avoids boot failures of a virtual machine when the last device in a PCI subsystem is assigned to a virtual machine on an AMD host system. The mlx. 4 driver must program the mlx. MAC addresses to listen to, including multicast addresses. Therefore, the mlx. MAC addresses. The driver used to perform updates to this list on the card by emptying the entire list and then programming in all of the addresses. Thus, whenever a user added or removed a multicast address or put the card into or out of promiscuous mode, the card's entire address list was re- written. This introduced a race condition, which resulted in a packet loss if a packet came in on an address the card should be listening to, but had not yet been reprogrammed to listen to. With this update, the driver no longer rewrites the entire list of trusted MAC addresses on the card but maintains a list of addresses that are currently programmed into the card. On address addition, only the new address is added to the end of the list, and on removal, only the to- be- removed address is removed from the list. The mlx. 4 card no longer experiences the described race condition and packets are no longer dropped in this scenario. Filesystem in Userspace (FUSE) did not implement scatter- gather direct I/O optimally. Consequently, the kernel had to process an extensive number of FUSE requests, which had a negative impact on system performance. This update applies a set of patches which improves internal request management for other features, such as readahead. FUSE direct I/O overhead has been significantly reduced to minimize negative effects on system performance. A previous kernel update introduced a bug that caused RAID0 and linear arrays larger than 4 TB to be truncated to 4 TB when using 0. The underlying source code has been modified so that 0. RAID0 and linear arrays larger than 4 TB are no longer truncated in the md RAID layer. A larger command descriptor block (CDB) is allocated for devices using Data Integrity Field (DIF) type 2 protection. The CDB was being freed in the sd_done() function, which resulted in a kernel panic if the command had to be retried in certain error recovery cases. With this update, the larger CDB is now freed in the sd_unprep_fn() function instead. This prevents the kernel panic from occurring. Previously, a use- after- free bug in the usbhid code caused a NULL pointer dereference. Consequent kernel memory corruption resulted in a kernel panic and could cause data loss. This update adds a NULL check to avoid these problems. A race condition could occur between page table sharing and virtual memory area (VMA) teardown. As a consequence, multiple "bad pmd" message warnings were displayed and "kernel BUG at mm/filemap. With this update, the VM_MAYSHARE flag is explicitly cleaned during the unmap_hugepage_range() call under the i_mmap_lock. This makes VMA ineligible for sharing and avoids the race condition. After using shared segments backed by huge pages, applications like databases and caches shut down correctly, with no crash. When I/O is issued through blk_execute_rq(), the blk_execute_rq_nowait() routine is called to perform various tasks. At first, the routine checks for a dead queue. Previously, however, if a dead queue was detected, the blk_execute_rq_nowait() function did not invoke the done() callback function. This resulted in blk_execute_rq() being unresponsive when waiting for completion, which had never been issued. To avoid such hangs, the rq- > end_io pointer is initialized to the done() callback before the queue state is verified. The Stream Control Transmission Protocol (SCTP) ipv. With this update, the source address is chosen from the routing table by taking this aspect into consideration.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. Archives
September 2016
Categories |