- 03 Apr, 2009 40 commits
-
-
Robert Love authored
This patch adds support for dynamically created Rx threads upon CPU hotplug events. There were existing synchronization problems that this patch attempts to resolve. The main problem had to do with fcoe_rcv() running in a different context than the hotplug notifications. This opened the possiblity that fcoe_rcv() would target a Rx thread for a skb. However, that thread could become NULL if the CPU was made offline. This patch uses the Rx queue's (a skb_queue) lock to protect the thread it's associated with and we use the 'thread' member of the fcoe_percpu_s to determine if the thread is ready to accept new skbs. The patch also attempts to do a better job of cleaning up, both if hotplug registration fails as well as when the module is removed. Contribution provided by Joe Eykholt <jeykholt@cisco.com> to fix incorrect use of __cpuinitdata. Signed-off-by:
Yi Zou <yi.zou@intel.com> Signed-off-by:
Robert Love <robert.w.love@intel.com> Signed-off-by:
James Bottomley <James.Bottomley@HansenPartnership.com>
-
Robert Love authored
Remove the hotplug creation of dev_stats, we allocate for all possible CPUs now when we allocate the lport. v2: Durring the 2.6.30 merge window, before these patches were comitted, 'percpu_ptr' was renamed 'per_cpu_ptr'. This latest update updates this patch for the name change. Signed-off-by:
-
Robert Love authored
Convert fcoe_percpu array to use the per-cpu variables that the kernel provides. Use the kernel's functions to access this structure. The cpu member of the fcoe_percpu_s is no longer needed, so this patch removes it too. Signed-off-by:
-
Robert Love authored
Currently the skb_queue is initialized every time the associated CPU goes online. This patch has libfcoe initializing the skb_queue for all possible CPUs when the module is loaded. This patch also re-orders some declarations in the fcoe_rcv() function so the structure declarations are grouped before the primitive declarations. Lastly, this patch converts all CPU indicies to use unsigned int since CPU indicies should not be negative. Signed-off-by:
-
Boaz Harrosh authored
Remove the creation of the symlink from the device to it's class. On modern systems this is already created by a udev rule and would WARN on load. On old systems it is not needed, none of the current osd user-mode tools use this link. Signed-off-by:
Boaz Harrosh <bharrosh@panasas.com> Signed-off-by:
-
Boaz Harrosh authored
A fix for a very serious and stupid bug in osd_initiator. It used to call blk_put_request() regardless of if it was from the end_io callback or if called after a sync execution. It should call the unlocked version __blk_put_request() instead. Also fixed is the remove of _abort_unexecuted_bios hack, and use of blk_end_request(,-ERROR,) to deallocate half baked requests. I've audited the code and it should be safe. Reported and Tested-by:
Xu Yang <onlyxuyang@qq.com> Signed-off-by:
-
James Bottomley authored
We have a problem with recovered error handling in that any command which goes down as BLOCK_PC but which returns a sense code of RECOVERED ERROR gets completed with -EIO. For actual SG_IO commands, this doesn't matter at all, since the error return code gets dropped in favour of req->errors which contain the SCSI completion code. However, if this command is part of the block system, then it will pay attention to the returned error code. In particularly if a SYNCHRONIZE CACHE from a barrier command completes with RECOVERED ERROR, the resulting -EIO on the barrier causes block to error the request and return it to the filesystem. Fix this by converting the -EIO for recovered error to zero, plus remove the printing of this from sd and sr so the message isn't double printed. Signed-off-by: -
Hannes Reinecke authored
aic79xx leaves timers inserted when ahd_init() (which inserts two timers at its very end) succeeds but ahd_pci_map_int() fails. In this case ahd->init_level gets incremented to 5 only when that function succeeds, but ahd_free() calls ahd_shutdown() only when ahd->init_level == 5, and ahd_shutdown() is where the timers get removed. Since the freeing of the IRQ is not controlled by ahd->init_level, we should increment init_level prior to calling ahd_pci_map_int(). Reported-by:
Jan Beulich <jbeulich@novell.com> Signed-off-by:
Hannes Reinecke <hare@suse.de> Signed-off-by:
-
Andrew Vasquez authored
Signed-off-by:
Andrew Vasquez <andrew.vasquez@qlogic.com> Signed-off-by:
-
Harish Zunjarrao authored
FLTDS provides FLT address in the byte address format, convert it to dword address for further use. Signed-off-by:
Harish Zunjarrao <harish.zunjarrao@qlogic.com> Signed-off-by:
-
Seokmann Ju authored
In addition to checking for potentially unnecessary iomem readX()/writeX() operations, a pci_channel_io_perm_failure should not trigger a full internal removal. Found during additional testing with pSeries blade systems. Signed-off-by:
Seokmann Ju <seokmann.ju@qlogic.com> Signed-off-by:
-
Andrew Vasquez authored
As updates will occur using low-level option-rom manipulation routines. Signed-off-by:
-
Andrew Vasquez authored
Firmware semantics changed for 24xx and above ISPs in their handling of the specified execution-throttle passed during firmware initialization. The original codes use of a theoretical maximum (0xffff, as carried over from earlier ISPs) could in fact act as a throttle in some circumstances. Now set the value based of the firmware's own 'resource' (exchange IOCBs) capabilities. Signed-off-by:
-
Andrew Vasquez authored
The original code to 'resize request-queues' based on iocb-counts and employed during early ISP23xx testing was too overly-pessimistic with regards to latencies in the firmware pulling requests. Recent ISPs can easily keep up processing a stream of commands from an abbreviated (effectively, half the original size) queue. Signed-off-by:
-
Andrew Vasquez authored
Rather than assuming a particular layout of the data. Applies to recent ISPs only. Signed-off-by:
-
Giridhar Malavali authored
When IRQs are shared by multiple controllers and if the first one to register does not disable the IRQ, then IRQ will be enabled for all other controllers by default, irrespective of their setting. With IRQF_DISABLED registration, the driver interrupt routine was called with interrupt enabled always. Disbaling the registration with IRQF_DISABLED, since driver code is re-entrant safe and all critical sections are guarded with interrupt safe locks. Signed-off-by:
Giridhar Malavali <giridhar.malavali@qlogic.com> Signed-off-by:
-
Joe Carnuccio authored
Interface allows for the update of onboard EDC firmware present on mezzanine ISP25xx type cards. Signed-off-by:
-
Lalit Chandivade authored
In handling the RMW semantics needed to update regions not falling on a sector boundary. Signed-off-by:
Lalit Chandivade <lalit.chandivade@qlogic.com> Signed-off-by:
-
Shyam Sundar authored
General cleanup of extraneous/legacy crud. Additional cleanups and Signed-off-by:
-
Lalit Chandivade authored
For ISP24xx and above the ISP-abort after flash update is not needed, as the only purpose it was serving was to update the boot code and firmware versions in the scsi_qla_host_t structure. Now an update of the versions will be done in the write-vpd path. Signed-off-by:
-
Lalit Chandivade authored
Signed-off-by:
-
Lalit Chandivade authored
Ensure that an ISP-abort has completed before performing any update. After the update do not wait for an ISP-abort completion, instead just wait until the ISP is reset. This avoids long delays due to waiting for loop ready in qla2x00_abort_isp(). Signed-off-by:
-
Joe Carnuccio authored
Given the low-level interface varies from one flash-part manufacturer to the next, the Flash-Access-Control (FAC) mailbox command makes the specific flash type transparent to the driver by encapsulating a basic set of accessor and update routines. Use these new routines where applicable by querying FAC opcode get-sector-size at init-time. Additional cleanups and Signed-off-by:
-
Andrew Vasquez authored
Since the routines can/will use resources such as devices and rports that aren't valid after midlayer tear-down, correct this potential race, by stopping the offending during the early stages of the remove() callback. Signed-off-by:
-
Andrew Vasquez authored
As all commands queued on the physical HBA should be aborted and returned to the upper-layers. Signed-off-by:
-
Andrew Vasquez authored
Reflects layout and format of latest specification. Signed-off-by:
-
Andrew Vasquez authored
Since in some circumstances, login-retries may be occuring in the background via the DPC routine. This race, in the inadvertant setting of the loop-id to 'NONE' breaks the existing retry logic. Signed-off-by:
-
Andrew Vasquez authored
Recent ISPs use this data to configure FCF information. Signed-off-by:
-
Andrew Vasquez authored
With recent ISPs loading firmware from flash, a flash-update to the firmware-image region with a follow-on reset will reload the new image. Original caching of data only made sense when firmware was bound with the driver. Signed-off-by:
-
Harish Zunjarrao authored
The pcihdr variable is used to find valid boot code image to get FC boot code versions from flash. The pcihdr variable should be byte aligned. Signed-off-by:
-
Giridhar Malavali authored
fce_size should be calculated based on the FCE_SIZE #define. Signed-off-by:
-
Giridhar Malavali authored
Earlier refactoring codes missed passing the proper vha structure and instead passed the 'hardware-descriptor' ha. Signed-off-by:
-
Giridhar Malavali authored
Post refactoring/multi-queue additions essentially eliminated the need for separate ISP24XX+ queuecommand as isp_ops contains a function pointer to the associated 'start_scsi()' operation. Signed-off-by:
-
Andrew Vasquez authored
Original code used an overabundance of indirect pointers to function helpers. Instead, the driver can exploit the immutable properties of a queue's ISP-association and ID, which are both known at queue initialization-time. Signed-off-by:
-
Brian King authored
Bump driver version to 1.0.5. Signed-off-by:
Brian King <brking@linux.vnet.ibm.com> Signed-off-by:
-
Brian King authored
The ibmvfc driver currently breaks the CRQ and essentially resets the entire virtual FC adapter, killing all outstanding ops to all attached targets, if an ADISC times out during target discover/rediscovery. This patch adds some code to cancel the ADISC if it times out, which prevents a single ADISC timeout from affecting the other devices attached to the fabric. Signed-off-by:
-
Brian King authored
Set show_host_maxframe_size so that maxframe_size gets exported in sysfs for the host. Signed-off-by:
-
Brian King authored
The ibmvfc driver really does not handle dynamically changing disc_threads. To change this dynamically would cause confusion in the driver regarding the number of event structs allocated. Fix this by simply not allowing disc_threads to be changed at runtime. Signed-off-by:
-
Brian King authored
This patch fixes a problem of possible dropped interrupts. Currently, the ibmvfc driver has a race condition where after ibmvfc_interrupt gets run, the platform code clears the interrupt. This can result in lost interrupts and, in worst case scenarios, result in command timeouts. Fix this by implementing a tasklet similar to what the ibmvscsi driver does so that interrupt processing is no longer done in the actual interrupt handler, which eliminates the race. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6Linus Torvalds authored
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6: Remove two unneeded exports and make two symbols static in fs/mpage.c Cleanup after commit 585d3bc0 Trim includes of fdtable.h Don't crap into descriptor table in binfmt_som Trim includes in binfmt_elf Don't mess with descriptor table in load_elf_binary() Get rid of indirect include of fs_struct.h New helper - current_umask() check_unsafe_exec() doesn't care about signal handlers sharing New locking/refcounting for fs_struct Take fs_struct handling to new file (fs/fs_struct.c) Get rid of bumping fs_struct refcount in pivot_root(2) Kill unsharing fs_struct in __set_personality()
-
