615 lines · plain
1Release Date : Thu Nov 16 15:32:35 EST 2006 -2 Sumant Patro <sumant.patro@lsi.com>3Current Version : 2.20.5.1 (scsi module), 2.20.2.6 (cmm module)4Older Version : 2.20.4.9 (scsi module), 2.20.2.6 (cmm module)5 61. Changes in Initialization to fix kdump failure.7 Send SYNC command on loading.8 This command clears the pending commands in the adapter9 and re-initialize its internal RAID structure.10 Without this change, megaraid driver either panics or fails to11 initialize the adapter during kdump's second kernel boot12 if there are pending commands or interrupts from other devices13 sharing the same IRQ.142. Authors email-id domain name changed from lsil.com to lsi.com.15 Also modified the MODULE_AUTHOR to megaraidlinux@lsi.com16 17Release Date : Fri May 19 09:31:45 EST 2006 - Seokmann Ju <sju@lsil.com>18Current Version : 2.20.4.9 (scsi module), 2.20.2.6 (cmm module)19Older Version : 2.20.4.8 (scsi module), 2.20.2.6 (cmm module)20 211. Fixed a bug in megaraid_init_mbox().22 Customer reported "garbage in file on x86_64 platform".23 Root Cause: the driver registered controllers as 64-bit DMA capable24 for those which are not support it.25 Fix: Made change in the function inserting identification mechanism26 identifying 64-bit DMA capable controllers.27 28 > -----Original Message-----29 > From: Vasily Averin [mailto:vvs@sw.ru]30 > Sent: Thursday, May 04, 2006 2:49 PM31 > To: linux-scsi@vger.kernel.org; Kolli, Neela; Mukker, Atul;32 > Ju, Seokmann; Bagalkote, Sreenivas;33 > James.Bottomley@SteelEye.com; devel@openvz.org34 > Subject: megaraid_mbox: garbage in file35 >36 > Hello all,37 >38 > I've investigated customers claim on the unstable work of39 > their node and found a40 > strange effect: reading from some files leads to the41 > "attempt to access beyond end of device" messages.42 >43 > I've checked filesystem, memory on the node, motherboard BIOS44 > version, but it45 > does not help and issue still has been reproduced by simple46 > file reading.47 >48 > Reproducer is simple:49 >50 > echo 0xffffffff >/proc/sys/dev/scsi/logging_level ;51 > cat /vz/private/101/root/etc/ld.so.cache >/tmp/ttt ;52 > echo 0 >/proc/sys/dev/scsi/logging53 >54 > It leads to the following messages in dmesg55 >56 > sd_init_command: disk=sda, block=871769260, count=2657 > sda : block=87176926058 > sda : reading 26/26 512 byte blocks.59 > scsi_add_timer: scmd: f79ed980, time: 7500, (c02b1420)60 > sd 0:1:0:0: send 0xf79ed980 sd 0:1:0:0:61 > command: Read (10): 28 00 33 f6 24 ac 00 00 1a 0062 > buffer = 0xf7cfb540, bufflen = 13312, done = 0xc0366b40,63 > queuecommand 0xc034401064 > leaving scsi_dispatch_cmnd()65 > scsi_delete_timer: scmd: f79ed980, rtn: 166 > sd 0:1:0:0: done 0xf79ed980 SUCCESS 0 sd 0:1:0:0:67 > command: Read (10): 28 00 33 f6 24 ac 00 00 1a 0068 > scsi host busy 1 failed 069 > sd 0:1:0:0: Notifying upper driver of completion (result 0)70 > sd_rw_intr: sda: res=0x071 > 26 sectors total, 13312 bytes done.72 > use_sg is 473 > attempt to access beyond end of device74 > sda6: rw=0, want=1044134458, limit=95140136775 > Buffer I/O error on device sda6, logical block 52206722876 > attempt to access beyond end of device77 782. When INQUIRY with EVPD bit set issued to the MegaRAID controller,79 system memory gets corrupted.80 Root Cause: MegaRAID F/W handle the INQUIRY with EVPD bit set81 incorrectly.82 Fix: MegaRAID F/W has fixed the problem and being process of release,83 soon. Meanwhile, driver will filter out the request.84 853. One member in the data structure of the driver leads to unaligned86 issue on 64-bit platform.87 Customer reported "kernel unaligned access address" issue when88 application communicates with MegaRAID HBA driver.89 Root Cause: in uioc_t structure, one of member had misaligned and it90 led system to display the error message.91 Fix: A patch submitted to community from following folk.92 93 > -----Original Message-----94 > From: linux-scsi-owner@vger.kernel.org95 > [mailto:linux-scsi-owner@vger.kernel.org] On Behalf Of Sakurai Hiroomi96 > Sent: Wednesday, July 12, 2006 4:20 AM97 > To: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org98 > Subject: Re: Help: strange messages from kernel on IA64 platform99 >100 > Hi,101 >102 > I saw same message.103 >104 > When GAM(Global Array Manager) is started, The following105 > message output.106 > kernel: kernel unaligned access to 0xe0000001fe1080d4,107 > ip=0xa000000200053371108 >109 > The uioc structure used by ioctl is defined by packed,110 > the allignment of each member are disturbed.111 > In a 64 bit structure, the allignment of member doesn't fit 64 bit112 > boundary. this causes this messages.113 > In a 32 bit structure, we don't see the message because the allinment114 > of member fit 32 bit boundary even if packed is specified.115 >116 > patch117 > I Add 32 bit dummy member to fit 64 bit boundary. I tested.118 > We confirmed this patch fix the problem by IA64 server.119 >120 > **************************************************************121 > ****************122 > --- linux-2.6.9/drivers/scsi/megaraid/megaraid_ioctl.h.orig123 > 2006-04-03 17:13:03.000000000 +0900124 > +++ linux-2.6.9/drivers/scsi/megaraid/megaraid_ioctl.h125 > 2006-04-03 17:14:09.000000000 +0900126 > @@ -132,6 +132,10 @@127 > /* Driver Data: */128 > void __user * user_data;129 > uint32_t user_data_len;130 > +131 > + /* 64bit alignment */132 > + uint32_t pad_0xBC;133 > +134 > mraid_passthru_t __user *user_pthru;135 >136 > mraid_passthru_t *pthru32;137 > **************************************************************138 > ****************139 140Release Date : Mon Apr 11 12:27:22 EST 2006 - Seokmann Ju <sju@lsil.com>141Current Version : 2.20.4.8 (scsi module), 2.20.2.6 (cmm module)142Older Version : 2.20.4.7 (scsi module), 2.20.2.6 (cmm module)143 1441. Fixed a bug in megaraid_reset_handler().145 Customer reported "Unable to handle kernel NULL pointer dereference146 at virtual address 00000000" when system goes to reset condition147 for some reason. It happened randomly.148 Root Cause: in the megaraid_reset_handler(), there is possibility not149 returning pending packets in the pend_list if there are multiple150 pending packets.151 Fix: Made the change in the driver so that it will return all packets152 in the pend_list.153 1542. Added change request.155 As found in the following URL, rmb() only didn't help the156 problem. I had to increase the loop counter to 0xFFFFFF. (6 F's)157 http://marc.theaimsgroup.com/?l=linux-scsi&m=110971060502497&w=2158 159 I attached a patch for your reference, too.160 Could you check and get this fix in your driver?161 162 Best Regards,163 Jun'ichi Nomura164 165Release Date : Fri Nov 11 12:27:22 EST 2005 - Seokmann Ju <sju@lsil.com>166Current Version : 2.20.4.7 (scsi module), 2.20.2.6 (cmm module)167Older Version : 2.20.4.6 (scsi module), 2.20.2.6 (cmm module)168 1691. Sorted out PCI IDs to remove megaraid support overlaps.170 Based on the patch from Daniel, sorted out PCI IDs along with171 character node name change from 'megadev' to 'megadev_legacy' to avoid172 conflict.173 ---174 Hopefully we'll be getting the build restriction zapped much sooner, 175 but we should also be thinking about totally removing the hardware 176 support overlap in the megaraid drivers.177 178 This patch pencils in a date of Feb 06 for this, and performs some 179 printk abuse in hope that existing legacy users might pick up on what's180 going on.181 182 Signed-off-by: Daniel Drake <dsd@gentoo.org>183 ---184 1852. Fixed a issue: megaraid always fails to reset handler.186 ---187 I found that the megaraid driver always fails to reset the188 adapter with the following message:189 megaraid: resetting the host...190 megaraid mbox: reset sequence completed successfully191 megaraid: fast sync command timed out192 megaraid: reservation reset failed193 when the "Cluster mode" of the adapter BIOS is enabled.194 So, whenever the reset occurs, the adapter goes to195 offline and just become unavailable.196 197 Jun'ichi Nomura [mailto:jnomura@mtc.biglobe.ne.jp]198 ---199 200Release Date : Mon Mar 07 12:27:22 EST 2005 - Seokmann Ju <sju@lsil.com>201Current Version : 2.20.4.6 (scsi module), 2.20.2.6 (cmm module)202Older Version : 2.20.4.5 (scsi module), 2.20.2.5 (cmm module)203 2041. Added IOCTL backward compatibility.205 Convert megaraid_mm driver to new compat_ioctl entry points.206 I don't have easy access to hardware, so only compile tested.207 - Signed-off-by:Andi Kleen <ak@muc.de>208 2092. megaraid_mbox fix: wrong order of arguments in memset()210 That, BTW, shows why cross-builds are useful-the only indication of211 problem had been a new warning showing up in sparse output on alpha212 build (number of exceeding 256 got truncated).213 - Signed-off-by: Al Viro214 <viro@parcelfarce.linux.theplanet.co.uk>215 2163. Convert pci_module_init to pci_register_driver217 Convert from pci_module_init to pci_register_driver218 (from:http://kernelnewbies.org/KernelJanitors/TODO)219 - Signed-off-by: Domen Puncer <domen@coderock.org>220 2214. Use the pre defined DMA mask constants from dma-mapping.h222 Use the DMA_{64,32}BIT_MASK constants from dma-mapping.h when calling223 pci_set_dma_mask() or pci_set_consistent_dma_mask(). See224 http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for more225 details.226 Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>227 Signed-off-by: Domen Puncer <domen@coderock.org>228 2295. Remove SSID checking for Dobson, Lindsay, and Verde based products.230 Checking the SSVID/SSID for controllers which have Dobson, Lindsay,231 and Verde is unnecessary because device ID has been assigned by LSI232 and it is unique value. So, all controllers with these IOPs have to be233 supported by the driver regardless SSVID/SSID.234 2356. Date Thu, 27 Jan 2005 04:31:09 +0100 236 From Herbert Poetzl <> 237 Subject RFC: assert_spin_locked() for 2.6 238 239 Greetings!240 241 overcautious programming will kill your kernel ;)242 ever thought about checking a spin_lock or even243 asserting that it must be held (maybe just for244 spinlock debugging?) ...245 246 there are several checks present in the kernel247 where somebody does a variation on the following:248 249 BUG_ON(!spin_is_locked(&some_lock));250 251 so what's wrong about that? nothing, unless you252 compile the code with CONFIG_DEBUG_SPINLOCK but 253 without CONFIG_SMP ... in which case the BUG()254 will kill your kernel ...255 256 maybe it's not advised to make such assertions, 257 but here is a solution which works for me ...258 (compile tested for sh, x86_64 and x86, boot/run259 tested for x86 only)260 261 best,262 Herbert263 264 - Herbert Poetzl <herbert@13thfloor.at>, Thu, 27 Jan 2005265 266Release Date : Thu Feb 03 12:27:22 EST 2005 - Seokmann Ju <sju@lsil.com>267Current Version : 2.20.4.5 (scsi module), 2.20.2.5 (cmm module)268Older Version : 2.20.4.4 (scsi module), 2.20.2.4 (cmm module)269 2701. Modified name of two attributes in scsi_host_template.271 On Wed, 2005-02-02 at 10:56 -0500, Ju, Seokmann wrote:272 > + .sdev_attrs = megaraid_device_attrs,273 > + .shost_attrs = megaraid_class_device_attrs,274 275 These are, perhaps, slightly confusing names.276 The terms device and class_device have well defined meanings in the277 generic device model, neither of which is what you mean here.278 Why not simply megaraid_sdev_attrs and megaraid_shost_attrs?279 280 Other than this, it looks fine to me too.281 282Release Date : Thu Jan 27 00:01:03 EST 2005 - Atul Mukker <atulm@lsil.com>283Current Version : 2.20.4.4 (scsi module), 2.20.2.5 (cmm module)284Older Version : 2.20.4.3 (scsi module), 2.20.2.4 (cmm module)285 2861. Bump up the version of scsi module due to its conflict.287 288Release Date : Thu Jan 21 00:01:03 EST 2005 - Atul Mukker <atulm@lsil.com>289Current Version : 2.20.4.3 (scsi module), 2.20.2.5 (cmm module)290Older Version : 2.20.4.2 (scsi module), 2.20.2.4 (cmm module)291 2921. Remove driver ioctl for logical drive to scsi address translation and293 replace with the sysfs attribute. To remove drives and change294 capacity, application shall now use the device attribute to get the295 logical drive number for a scsi device. For adding newly created296 logical drives, class device attribute would be required to uniquely297 identify each controller.298 - Atul Mukker <atulm@lsil.com>299 300 "James, I've been thinking about this a little more, and you may be on301 to something here. Let each driver add files as such:"302 303 - Matt Domsch <Matt_Domsch@dell.com>, 12.15.2004304 linux-scsi mailing list305 306 307 "Then, if you simply publish your LD number as an extra parameter of308 the device, you can look through /sys to find it."309 310 - James Bottomley <James.Bottomley@SteelEye.com>, 01.03.2005311 linux-scsi mailing list312 313 314 "I don't see why not ... it's your driver, you can publish whatever315 extra information you need as scsi_device attributes; that was one of316 the designs of the extensible attribute system."317 318 - James Bottomley <James.Bottomley@SteelEye.com>, 01.06.2005319 linux-scsi mailing list320 3212. Add AMI megaraid support - Brian King <brking@charter.net>322 PCI_VENDOR_ID_AMI, PCI_DEVICE_ID_AMI_MEGARAID3,323 PCI_VENDOR_ID_AMI, PCI_SUBSYS_ID_PERC3_DC,324 3253. Make some code static - Adrian Bunk <bunk@stusta.de>326 Date: Mon, 15 Nov 2004 03:14:57 +0100327 328 The patch below makes some needlessly global code static.329 -wait_queue_head_t wait_q;330 +static wait_queue_head_t wait_q;331 332 Signed-off-by: Adrian Bunk <bunk@stusta.de>333 3344. Added NEC ROMB support - NEC MegaRAID PCI Express ROMB controller335 PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_MEGARAID_NEC_ROMB_2E,336 PCI_SUBSYS_ID_NEC, PCI_SUBSYS_ID_MEGARAID_NEC_ROMB_2E,337 3385. Fixed Tape drive issue : For any Direct CDB command to physical device339 including tape, timeout value set by driver was 10 minutes. With this 340 value, most of command will return within timeout. However, for those341 command like ERASE or FORMAT, it takes more than an hour depends on342 capacity of the device and the command could be terminated before it 343 completes.344 To address this issue, the 'timeout' field in the DCDB command will 345 have NO TIMEOUT (i.e., 4) value as its timeout on DCDB command.346 347 348 349Release Date : Thu Dec 9 19:10:23 EST 2004350 - Sreenivas Bagalkote <sreenib@lsil.com>351 352Current Version : 2.20.4.2 (scsi module), 2.20.2.4 (cmm module)353Older Version : 2.20.4.1 (scsi module), 2.20.2.3 (cmm module)354 355i. Introduced driver ioctl that returns scsi address for a given ld.356 357 "Why can't the existing sysfs interfaces be used to do this?"358 - Brian King (brking@us.ibm.com)359 360 "I've looked into solving this another way, but I cannot see how361 to get this driver-private mapping of logical drive number-> HCTL362 without putting code something like this into the driver."363 364 "...and by providing a mapping a function to userspace, the driver365 is free to change its mapping algorithm in the future if necessary .."366 - Matt Domsch (Matt_Domsch@dell.com)367 368Release Date : Thu Dec 9 19:02:14 EST 2004 - Sreenivas Bagalkote <sreenib@lsil.com>369 370Current Version : 2.20.4.1 (scsi module), 2.20.2.3 (cmm module)371Older Version : 2.20.4.1 (scsi module), 2.20.2.2 (cmm module)372 373i. Fix a bug in kioc's dma buffer deallocation374 375Release Date : Thu Nov 4 18:24:56 EST 2004 - Sreenivas Bagalkote <sreenib@lsil.com>376 377Current Version : 2.20.4.1 (scsi module), 2.20.2.2 (cmm module)378Older Version : 2.20.4.0 (scsi module), 2.20.2.1 (cmm module)379 380i. Handle IOCTL cmd timeouts more properly.381 382ii. pci_dma_sync_{sg,single}_for_cpu was introduced into megaraid_mbox383 incorrectly (instead of _for_device). Changed to appropriate 384 pci_dma_sync_{sg,single}_for_device.385 386Release Date : Wed Oct 06 11:15:29 EDT 2004 - Sreenivas Bagalkote <sreenib@lsil.com>387Current Version : 2.20.4.0 (scsi module), 2.20.2.1 (cmm module)388Older Version : 2.20.4.0 (scsi module), 2.20.2.0 (cmm module)389 390i. Remove CONFIG_COMPAT around register_ioctl32_conversion391 392Release Date : Mon Sep 27 22:15:07 EDT 2004 - Atul Mukker <atulm@lsil.com>393Current Version : 2.20.4.0 (scsi module), 2.20.2.0 (cmm module)394Older Version : 2.20.3.1 (scsi module), 2.20.2.0 (cmm module)395 396i. Fix data corruption. Because of a typo in the driver, the IO packets397 were wrongly shared by the ioctl path. This causes a whole IO command398 to be replaced by an incoming ioctl command.399 400Release Date : Tue Aug 24 09:43:35 EDT 2004 - Atul Mukker <atulm@lsil.com>401Current Version : 2.20.3.1 (scsi module), 2.20.2.0 (cmm module)402Older Version : 2.20.3.0 (scsi module), 2.20.2.0 (cmm module)403 404i. Function reordering so that inline functions are defined before they405 are actually used. It is now mandatory for GCC 3.4.1 (current stable)406 407 Declare some heavy-weight functions to be non-inlined,408 megaraid_mbox_build_cmd, megaraid_mbox_runpendq,409 megaraid_mbox_prepare_pthru, megaraid_mbox_prepare_epthru,410 megaraid_busywait_mbox411 412 - Andrew Morton, 08.19.2004413 linux-scsi mailing list414 415 "Something else to clean up after inclusion: every instance of an416 inline function is actually rendered as a full function call, because417 the function is always used before it is defined. Atul, please418 re-arrange the code to eliminate the need for most (all) of the419 function prototypes at the top of each file, and define (not just420 declare with a prototype) each inline function before its first use"421 422 - Matt Domsch <Matt_Domsch@dell.com>, 07.27.2004423 linux-scsi mailing list424 425 426ii. Display elapsed time (countdown) while waiting for FW to boot.427 428iii. Module compilation reorder in Makefile so that unresolved symbols do429 not occur when driver is compiled non-modular.430 431 Patrick J. LoPresti <patl@users.sourceforge.net>, 8.22.2004432 linux-scsi mailing list433 434 435Release Date : Thu Aug 19 09:58:33 EDT 2004 - Atul Mukker <atulm@lsil.com>436Current Version : 2.20.3.0 (scsi module), 2.20.2.0 (cmm module)437Older Version : 2.20.2.0 (scsi module), 2.20.1.0 (cmm module)438 439i. When copying the mailbox packets, copy only first 14 bytes (for 32-bit440 mailboxes) and only first 22 bytes (for 64-bit mailboxes). This is to441 avoid getting the stale values for busy bit. We want to set the busy442 bit just before issuing command to the FW.443 444ii. In the reset handling, if the reset command is not owned by the445 driver, do not (wrongly) print information for the "attached" driver446 packet.447 448iii. Have extended wait when issuing command in synchronous mode. This is449 required for the cases where the option ROM is disabled and there is450 no BIOS to start the controller. The FW starts to boot after receiving451 the first command from the driver. The current driver has 1 second452 timeout for the synchronous commands, which is far less than what is453 actually required. We now wait up to MBOX_RESET_TIME (180 seconds) for454 FW boot process.455 456iv. In megaraid_mbox_product_info, clear the mailbox contents completely457 before preparing the command for inquiry3. This is to ensure that the458 FW does not get junk values in the command.459 460v. Do away with the redundant LSI_CONFIG_COMPAT redefinition for461 CONFIG_COMPAT. Replace <asm/ioctl32.h> with <linux/ioctl32.h>462 463 - James Bottomley <James.Bottomley@SteelEye.com>, 08.17.2004464 linux-scsi mailing list465 466vi. Add support for 64-bit applications. Current drivers assume only467 32-bit applications, even on 64-bit platforms. Use the "data" and468 "buffer" fields of the mimd_t structure, instead of embedded 32-bit469 addresses in application mailbox and passthru structures.470 471vii. Move the function declarations for the management module from472 megaraid_mm.h to megaraid_mm.c473 474 - Andrew Morton, 08.19.2004475 linux-scsi mailing list476 477viii. Change default values for MEGARAID_NEWGEN, MEGARAID_MM, and478 MEGARAID_MAILBOX to 'n' in Kconfig.megaraid479 480 - Andrew Morton, 08.19.2004481 linux-scsi mailing list482 483ix. replace udelay with msleep484 485x. Typos corrected in comments and whitespace adjustments, explicit486 grouping of expressions.487 488 489Release Date : Fri Jul 23 15:22:07 EDT 2004 - Atul Mukker <atulm@lsil.com>490Current Version : 2.20.2.0 (scsi module), 2.20.1.0 (cmm module)491Older Version : 2.20.1.0 (scsi module), 2.20.0.0 (cmm module)492 493i. Add PCI ids for Acer ROMB 2E solution494 495ii. Add PCI ids for I4496 497iii. Typo corrected for subsys id for megaraid sata 300-4x498 499iv. Remove yield() while mailbox handshake in synchronous commands500 501 502 "My other main gripe is things like this:503 504 + // wait for maximum 1 second for status to post505 + for (i = 0; i < 40000; i++) {506 + if (mbox->numstatus != 0xFF) break;507 + udelay(25); yield();508 + }509 510 which litter the driver. Use of yield() in drivers is deprecated."511 512 - James Bottomley <James.Bottomley@SteelEye.com>, 07.14.2004513 linux-scsi mailing list514 515v. Remove redundant __megaraid_busywait_mbox routine516 517vi. Fix bug in the management module, which causes a system lockup when the518 IO module is loaded and then unloaded, followed by executing any519 management utility. The current version of management module does not520 handle the adapter unregister properly.521 522 Specifically, it still keeps a reference to the unregistered523 controllers. To avoid this, the static array adapters has been524 replaced by a dynamic list, which gets updated every time an adapter525 is added or removed.526 527 Also, during unregistration of the IO module, the resources are528 now released in the exact reverse order of the allocation time529 sequence.530 531 532Release Date : Fri Jun 25 18:58:43 EDT 2004 - Atul Mukker <atulm@lsil.com>533Current Version : 2.20.1.0534Older Version : megaraid 2.20.0.1535 536i. Stale list pointer in adapter causes kernel panic when module537 megaraid_mbox is unloaded538 539 540Release Date : Thu Jun 24 20:37:11 EDT 2004 - Atul Mukker <atulm@lsil.com>541Current Version : 2.20.0.1542Older Version : megaraid 2.20.0.00543 544i. Modules are not 'y' by default, but depend on current definition of545 SCSI & PCI.546 547ii. Redundant structure mraid_driver_t removed.548 549iii. Miscellaneous indentation and goto/label fixes.550 - Christoph Hellwig <hch@infradead.org>, 06.24.2004 linux-scsi551 552iv. scsi_host_put(), do just before completing HBA shutdown.553 554 555 556Release Date : Mon Jun 21 19:53:54 EDT 2004 - Atul Mukker <atulm@lsil.com>557Current Version : 2.20.0.0558Older Version : megaraid 2.20.0.rc2 and 2.00.3559 560i. Independent module to interact with userland applications and561 multiplex command to low level RAID module(s).562 563 "Shared code in a third module, a "library module", is an acceptable564 solution. modprobe automatically loads dependent modules, so users565 running "modprobe driver1" or "modprobe driver2" would automatically566 load the shared library module."567 568 - Jeff Garzik <jgarzik@pobox.com> 02.25.2004 LKML569 570 "As Jeff hinted, if your userspace<->driver API is consistent between571 your new MPT-based RAID controllers and your existing megaraid driver,572 then perhaps you need a single small helper module (lsiioctl or some573 better name), loaded by both mptraid and megaraid automatically, which574 handles registering the /dev/megaraid node dynamically. In this case,575 both mptraid and megaraid would register with lsiioctl for each576 adapter discovered, and lsiioctl would essentially be a switch,577 redirecting userspace tool ioctls to the appropriate driver."578 579 - Matt Domsch <Matt_Domsch@dell.com> 02.25.2004 LKML580 581ii. Remove C99 initializations from pci_device id.582 583 "pci_id_table_g would be much more readable when not using C99584 initializers.585 PCI table doesn't change, there's lots of users that prefer the more586 readable variant. And it's really far less and much easier to grok587 lines without C99 initializers."588 589 - Christoph Hellwig <hch@infradead.org>, 05.28.2004 linux-scsi590 591iii. Many fixes as suggested by Christoph Hellwig <hch@infradead.org> on592 linux-scsi, 05.28.2004593 594iv. We now support up to 32 parallel ioctl commands instead of current 1.595 There is a conscious effort to let memory allocation not fail for ioctl596 commands.597 598v. Do away with internal memory management. Use pci_pool_(create|alloc)599 instead.600 601vi. Kill tasklet when unloading the driver.602 603vii. Do not use "host_lock', driver has fine-grain locks now to protect all604 data structures.605 606viii. Optimize the build scatter-gather list routine. The callers already607 know the data transfer address and length.608 609ix. Better implementation of error handling and recovery. Driver now610 performs extended errors recovery for instances like scsi cable pull.611 612x. Disassociate the management commands with an overlaid scsi command.613 Driver now treats the management packets as special packets and has a614 dedicated callback routine.615