1 Release Date : Thu Nov 16 15:32:35 EST 2006
2 Sumant Patro <s
3 Current Version : 2.20.5.1 (scsi module), 2.20
4 Older Version : 2.20.4.9 (scsi module), 2.20
5
6 1. Changes in Initialization to fix kdump
7 Send SYNC command on loading.
8 This command clears the pending comman
9 and re-initialize its internal RAID st
10 Without this change, megaraid driver e
11 initialize the adapter during kdump's
12 if there are pending commands or inter
13 sharing the same IRQ.
14 2. Authors email-id domain name changed f
15 Also modified the MODULE_AUTHOR to meg
16
17 Release Date : Fri May 19 09:31:45 EST 2006<
18 Current Version : 2.20.4.9 (scsi module), 2.20
19 Older Version : 2.20.4.8 (scsi module), 2.20
20
21 1. Fixed a bug in megaraid_init_mbox().
22 Customer reported "garbage in file on
23 Root Cause: the driver registered cont
24 for those which are not support it.
25 Fix: Made change in the function inser
26 identifying 64-bit DMA capable control
27
28 > -----Original Message-----
29 > From: Vasily Averin [mailto:vvs@sw.r
30 > Sent: Thursday, May 04, 2006 2:49 PM
31 > To: linux-scsi@vger.kernel.org; Koll
32 > Ju, Seokmann; Bagalkote, Sreenivas;
33 > James.Bottomley@SteelEye.com; devel@
34 > Subject: megaraid_mbox: garbage in f
35 >
36 > Hello all,
37 >
38 > I've investigated customers claim on
39 > their node and found a
40 > strange effect: reading from some fi
41 > "attempt to access beyond end of de
42 >
43 > I've checked filesystem, memory on t
44 > version, but it
45 > does not help and issue still has be
46 > file reading.
47 >
48 > Reproducer is simple:
49 >
50 > echo 0xffffffff >/proc/sys/dev/scsi/
51 > cat /vz/private/101/root/etc/ld.so.c
52 > echo 0 >/proc/sys/dev/scsi/logging
53 >
54 > It leads to the following messages i
55 >
56 > sd_init_command: disk=sda, block=871
57 > sda : block=871769260
58 > sda : reading 26/26 512 byte blocks.
59 > scsi_add_timer: scmd: f79ed980, time
60 > sd 0:1:0:0: send 0xf79ed980
61 > command: Read (10): 28 00 33
62 > buffer = 0xf7cfb540, bufflen = 13312
63 > queuecommand 0xc0344010
64 > leaving scsi_dispatch_cmnd()
65 > scsi_delete_timer: scmd: f79ed980, r
66 > sd 0:1:0:0: done 0xf79ed980 SUCCESS
67 > command: Read (10): 28 00 33
68 > scsi host busy 1 failed 0
69 > sd 0:1:0:0: Notifying upper driver o
70 > sd_rw_intr: sda: res=0x0
71 > 26 sectors total, 13312 bytes done.
72 > use_sg is 4
73 > attempt to access beyond end of devi
74 > sda6: rw=0, want=1044134458, limit=9
75 > Buffer I/O error on device sda6, log
76 > attempt to access beyond end of devi
77
78 2. When INQUIRY with EVPD bit set issued
79 system memory gets corrupted.
80 Root Cause: MegaRAID F/W handle the IN
81 incorrectly.
82 Fix: MegaRAID F/W has fixed the proble
83 soon. Meanwhile, driver will filter ou
84
85 3. One member in the data structure of th
86 issue on 64-bit platform.
87 Customer reported "kernel unaligned ac
88 application communicates with MegaRAID
89 Root Cause: in uioc_t structure, one o
90 led system to display the error messag
91 Fix: A patch submitted to community fr
92
93 > -----Original Message-----
94 > From: linux-scsi-owner@vger.kernel.o
95 > [mailto:linux-scsi-owner@vger.kernel
96 > Sent: Wednesday, July 12, 2006 4:20
97 > To: linux-scsi@vger.kernel.org; linu
98 > Subject: Re: Help: strange messages
99 >
100 > Hi,
101 >
102 > I saw same message.
103 >
104 > When GAM(Global Array Manager) is st
105 > message output.
106 > kernel: kernel unaligned access to 0
107 > ip=0xa000000200053371
108 >
109 > The uioc structure used by ioctl is
110 > the allignment of each member are di
111 > In a 64 bit structure, the allignmen
112 > boundary. this causes this messages.
113 > In a 32 bit structure, we don't see
114 > of member fit 32 bit boundary even i
115 >
116 > patch
117 > I Add 32 bit dummy member to fit 64
118 > We confirmed this patch fix the prob
119 >
120 > ************************************
121 > ****************
122 > --- linux-2.6.9/drivers/scsi/megarai
123 > 2006-04-03 17:13:03.000000000 +0900
124 > +++ linux-2.6.9/drivers/scsi/megarai
125 > 2006-04-03 17:14:09.000000000 +0900
126 > @@ -132,6 +132,10 @@
127 > /* Driver Data: */
128 > void __user * use
129 > uint32_t use
130 > +
131 > + /* 64bit alignment */
132 > + uint32_t pad
133 > +
134 > mraid_passthru_t __u
135 >
136 > mraid_passthru_t *pt
137 > ************************************
138 > ****************
139
140 Release Date : Mon Apr 11 12:27:22 EST 2006<
141 Current Version : 2.20.4.8 (scsi module), 2.20
142 Older Version : 2.20.4.7 (scsi module), 2.20
143
144 1. Fixed a bug in megaraid_reset_handler(
145 Customer reported "Unable to handle ke
146 at virtual address 00000000" when syst
147 for some reason. It happened randomly.
148 Root Cause: in the megaraid_reset_hand
149 returning pending packets in the pend_
150 pending packets.
151 Fix: Made the change in the driver so
152 in the pend_list.
153
154 2. Added change request.
155 As found in the following URL, rmb() o
156 problem. I had to increase the loop co
157 http://marc.theaimsgroup.com/?l=linux-
158
159 I attached a patch for your reference,
160 Could you check and get this fix in yo
161
162 Best Regards,
163 Jun'ichi Nomura
164
165 Release Date : Fri Nov 11 12:27:22 EST 2005<
166 Current Version : 2.20.4.7 (scsi module), 2.20
167 Older Version : 2.20.4.6 (scsi module), 2.20
168
169 1. Sorted out PCI IDs to remove megaraid
170 Based on the patch from Daniel, sorted
171 character node name change from 'megad
172 conflict.
173 ---
174 Hopefully we'll be getting the build r
175 but we should also be thinking about t
176 support overlap in the megaraid driver
177
178 This patch pencils in a date of Feb 06
179 printk abuse in hope that existing leg
180 going on.
181
182 Signed-off-by: Daniel Drake <dsd@gentoo
183 ---
184
185 2. Fixed a issue: megaraid always fails t
186 ---
187 I found that the megaraid driver alway
188 adapter with the following message:
189 megaraid: resetting the host..
190 megaraid mbox: reset sequence
191 megaraid: fast sync command ti
192 megaraid: reservation reset fa
193 when the "Cluster mode" of the adapter
194 So, whenever the reset occurs, the ada
195 offline and just become unavailable.
196
197 Jun'ichi Nomura [mailto:jnomura@mtc.bi
198 ---
199
200 Release Date : Mon Mar 07 12:27:22 EST 2005<
201 Current Version : 2.20.4.6 (scsi module), 2.20
202 Older Version : 2.20.4.5 (scsi module), 2.20
203
204 1. Added IOCTL backward compatibility.
205 Convert megaraid_mm driver to new comp
206 I don't have easy access to hardware,
207 - Signed-off-by:Andi Kleen <ak@
208
209 2. megaraid_mbox fix: wrong order of argu
210 That, BTW, shows why cross-builds are
211 problem had been a new warning showing
212 build (number of exceeding 256 got tru
213 - Signed-off-by: Al Viro
214 <viro@parcelfarce.linux.theplan
215
216 3. Convert pci_module_init to pci_registe
217 Convert from pci_module_init to pci_re
218 (from:http://kernelnewbies.org/KernelJ
219 - Signed-off-by: Domen Puncer <
220
221 4. Use the pre defined DMA mask constants
222 Use the DMA_{64,32}BIT_MASK constants
223 pci_set_dma_mask() or pci_set_consiste
224 http://marc.theaimsgroup.com/?t=108001
225 details.
226 Signed-off-by: Tobias Klauser <
227 Signed-off-by: Domen Puncer <do
228
229 5. Remove SSID checking for Dobson, Linds
230 Checking the SSVID/SSID for controller
231 and Verde is unnecessary because devic
232 and it is unique value. So, all contro
233 supported by the driver regardless SSV
234
235 6. Date Thu, 27 Jan 2005 04:31:09 +0100
236 From Herbert Poetzl <>
237 Subject RFC: assert_spin_locked() for
238
239 Greetings!
240
241 overcautious programming will kill you
242 ever thought about checking a spin_loc
243 asserting that it must be held (maybe
244 spinlock debugging?) ...
245
246 there are several checks present in th
247 where somebody does a variation on the
248
249 BUG_ON(!spin_is_locked(&some_lock));
250
251 so what's wrong about that? nothing, u
252 compile the code with CONFIG_DEBUG_SPI
253 without CONFIG_SMP ... in which case t
254 will kill your kernel ...
255
256 maybe it's not advised to make such as
257 but here is a solution which works for
258 (compile tested for sh, x86_64 and x86
259 tested for x86 only)
260
261 best,
262 Herbert
263
264 - Herbert Poetzl <herbert@13thf
265
266 Release Date : Thu Feb 03 12:27:22 EST 2005<
267 Current Version : 2.20.4.5 (scsi module), 2.20
268 Older Version : 2.20.4.4 (scsi module), 2.20
269
270 1. Modified name of two attributes in scs
271 On Wed, 2005-02-02 at 10:56 -0500, Ju,
272 > + .sdev_attrs
273 > + .shost_attrs
274
275 These are, perhaps, slightly confusing
276 The terms device and class_device have
277 generic device model, neither of which
278 Why not simply megaraid_sdev_attrs and
279
280 Other than this, it looks fine to me t
281
282 Release Date : Thu Jan 27 00:01:03 EST 2005<
283 Current Version : 2.20.4.4 (scsi module), 2.20
284 Older Version : 2.20.4.3 (scsi module), 2.20
285
286 1. Bump up the version of scsi module due
287
288 Release Date : Thu Jan 21 00:01:03 EST 2005<
289 Current Version : 2.20.4.3 (scsi module), 2.20
290 Older Version : 2.20.4.2 (scsi module), 2.20
291
292 1. Remove driver ioctl for logical drive
293 replace with the sysfs attribute. To r
294 capacity, application shall now use th
295 logical drive number for a scsi device
296 logical drives, class device attribute
297 identify each controller.
298 - Atul Mukker <atulm@lsil.com>
299
300 "James, I've been thinking about this
301 to something here. Let each driver add
302
303 - Matt Domsch <Matt_Domsch@dell
304 linux-scsi mailing list
305
306
307 "Then, if you simply publish your LD n
308 the device, you can look through /sys
309
310 - James Bottomley <James.Bottom
311 linux-scsi mailing list
312
313
314 "I don't see why not ... it's your dri
315 extra information you need as scsi_dev
316 the designs of the extensible attribut
317
318 - James Bottomley <James.Bottom
319 linux-scsi mailing list
320
321 2. Add AMI megaraid support - Brian King <
322 PCI_VENDOR_ID_AMI, PCI_DEVICE_
323 PCI_VENDOR_ID_AMI, PCI_SUBSYS_
324
325 3. Make some code static - Adrian Bunk <bu
326 Date: Mon, 15 Nov 2004 03:14:57 +010
327
328 The patch below makes some needlessly
329 -wait_queue_head_t wait_q;
330 +static wait_queue_head_t wait_q;
331
332 Signed-off-by: Adrian Bunk <bunk@stusta
333
334 4. Added NEC ROMB support - NEC MegaRAID
335 PCI_VENDOR_ID_LSI_LOGIC, PCI_D
336 PCI_SUBSYS_ID_NEC, PCI_SUBSYS_
337
338 5. Fixed Tape drive issue : For any Direc
339 including tape, timeout value set by d
340 value, most of command will return wit
341 command like ERASE or FORMAT, it takes
342 capacity of the device and the command
343 completes.
344 To address this issue, the 'timeout' f
345 have NO TIMEOUT (i.e., 4) value as its
346
347
348
349 Release Date : Thu Dec 9 19:10:23 EST 2004
350 - Sreenivas Bagalkote <sreenib@lsil.com
351
352 Current Version : 2.20.4.2 (scsi module), 2.20
353 Older Version : 2.20.4.1 (scsi module), 2.20
354
355 i. Introduced driver ioctl that returns s
356
357 "Why can't the existing sysfs interfac
358 - Brian King (brking@us.ibm.co
359
360 "I've looked into solving this another
361 to get this driver-private mapping of
362 without putting code something like th
363
364 "...and by providing a mapping a funct
365 is free to change its mapping algorith
366 - Matt Domsch (Matt_Domsch@del
367
368 Release Date : Thu Dec 9 19:02:14 EST 2004<
369
370 Current Version : 2.20.4.1 (scsi module), 2.20
371 Older Version : 2.20.4.1 (scsi module), 2.20
372
373 i. Fix a bug in kioc's dma buffer dealloc
374
375 Release Date : Thu Nov 4 18:24:56 EST 2004<
376
377 Current Version : 2.20.4.1 (scsi module), 2.20
378 Older Version : 2.20.4.0 (scsi module), 2.20
379
380 i. Handle IOCTL cmd timeouts more properl
381
382 ii. pci_dma_sync_{sg,single}_for_cpu was i
383 incorrectly (instead of _for_device).
384 pci_dma_sync_{sg,single}_for_device.
385
386 Release Date : Wed Oct 06 11:15:29 EDT 2004<
387 Current Version : 2.20.4.0 (scsi module), 2.20
388 Older Version : 2.20.4.0 (scsi module), 2.20
389
390 i. Remove CONFIG_COMPAT around register_i
391
392 Release Date : Mon Sep 27 22:15:07 EDT 2004<
393 Current Version : 2.20.4.0 (scsi module), 2.20
394 Older Version : 2.20.3.1 (scsi module), 2.20
395
396 i. Fix data corruption. Because of a typo
397 were wrongly shared by the ioctl path.
398 to be replaced by an incoming ioctl co
399
400 Release Date : Tue Aug 24 09:43:35 EDT 2004<
401 Current Version : 2.20.3.1 (scsi module), 2.20
402 Older Version : 2.20.3.0 (scsi module), 2.20
403
404 i. Function reordering so that inline fun
405 are actually used. It is now mandatory
406
407 Declare some heavy-weight functions to
408 megaraid_mbox_build_cmd, megaraid_mbox
409 megaraid_mbox_prepare_pthru, megaraid_
410 megaraid_busywait_mbox
411
412 - Andrew Morton, 08.19.2004
413 linux-scsi mailing list
414
415 "Something else to clean up after incl
416 inline function is actually rendered a
417 the function is always used before it
418 re-arrange the code to eliminate the n
419 function prototypes at the top of each
420 declare with a prototype) each inline
421
422 - Matt Domsch <Matt_Domsch@dell
423 linux-scsi mailing list
424
425
426 ii. Display elapsed time (countdown) while
427
428 iii. Module compilation reorder in Makefile
429 not occur when driver is compiled non-
430
431 Patrick J. LoPresti <patl@users
432 linux-scsi mailing list
433
434
435 Release Date : Thu Aug 19 09:58:33 EDT 2004<
436 Current Version : 2.20.3.0 (scsi module), 2.20
437 Older Version : 2.20.2.0 (scsi module), 2.20
438
439 i. When copying the mailbox packets, copy
440 mailboxes) and only first 22 bytes (fo
441 avoid getting the stale values for bus
442 bit just before issuing command to the
443
444 ii. In the reset handling, if the reset co
445 driver, do not (wrongly) print informa
446 packet.
447
448 iii. Have extended wait when issuing comman
449 required for the cases where the optio
450 no BIOS to start the controller. The F
451 the first command from the driver. The
452 timeout for the synchronous commands,
453 actually required. We now wait up to M
454 FW boot process.
455
456 iv. In megaraid_mbox_product_info, clear t
457 before preparing the command for inqui
458 FW does not get junk values in the com
459
460 v. Do away with the redundant LSI_CONFIG_
461 CONFIG_COMPAT. Replace <asm/ioctl32.h>
462
463 - James Bottomley <James.Bottom
464 linux-scsi mailing list
465
466 vi. Add support for 64-bit applications. C
467 32-bit applications, even on 64-bit pl
468 "buffer" fields of the mimd_t structur
469 addresses in application mailbox and p
470
471 vii. Move the function declarations for the
472 megaraid_mm.h to megaraid_mm.c
473
474 - Andrew Morton, 08.19.2004
475 linux-scsi mailing list
476
477 viii. Change default values for MEGARAID_NEW
478 MEGARAID_MAILBOX to 'n' in Kconfig.meg
479
480 - Andrew Morton, 08.19.2004
481 linux-scsi mailing list
482
483 ix. replace udelay with msleep
484
485 x. Typos corrected in comments and whites
486 grouping of expressions.
487
488
489 Release Date : Fri Jul 23 15:22:07 EDT 2004<
490 Current Version : 2.20.2.0 (scsi module), 2.20
491 Older Version : 2.20.1.0 (scsi module), 2.20
492
493 i. Add PCI ids for Acer ROMB 2E solution
494
495 ii. Add PCI ids for I4
496
497 iii. Typo corrected for subsys id for megar
498
499 iv. Remove yield() while mailbox handshake
500
501
502 "My other main gripe is things like th
503
504 + // wait for maximum 1 second f
505 + for (i = 0; i < 40000; i++) {
506 + if (mbox->numstatus !=
507 + udelay(25); yield();
508 + }
509
510 which litter the driver. Use of yield
511
512 - James Bottomley <James.Bottom
513 linux-scsi mailing list
514
515 v. Remove redundant __megaraid_busywait_m
516
517 vi. Fix bug in the management module, whic
518 IO module is loaded and then unloaded,
519 management utility. The current versio
520 handle the adapter unregister properly
521
522 Specifically, it still keeps a referen
523 controllers. To avoid this, the static
524 replaced by a dynamic list, which gets
525 is added or removed.
526
527 Also, during unregistration of the IO
528 now released in the exact reverse orde
529 sequence.
530
531
532 Release Date : Fri Jun 25 18:58:43 EDT 2004<
533 Current Version : 2.20.1.0
534 Older Version : megaraid 2.20.0.1
535
536 i. Stale list pointer in adapter causes k
537 megaraid_mbox is unloaded
538
539
540 Release Date : Thu Jun 24 20:37:11 EDT 2004<
541 Current Version : 2.20.0.1
542 Older Version : megaraid 2.20.0.00
543
544 i. Modules are not 'y' by default, but de
545 SCSI & PCI.
546
547 ii. Redundant structure mraid_driver_t rem
548
549 iii. Miscellaneous indentation and goto/lab
550 - Christoph Hellwig <hch@infrad
551
552 iv. scsi_host_put(), do just before comple
553
554
555
556 Release Date : Mon Jun 21 19:53:54 EDT 2004<
557 Current Version : 2.20.0.0
558 Older Version : megaraid 2.20.0.rc2 and 2.00
559
560 i. Independent module to interact with us
561 multiplex command to low level RAID mo
562
563 "Shared code in a third module, a "lib
564 solution. modprobe automatically loads
565 running "modprobe driver1" or "modprob
566 load the shared library module."
567
568 - Jeff Garzik <jgarzik@pobox.co
569
570 "As Jeff hinted, if your userspace<->d
571 your new MPT-based RAID controllers an
572 then perhaps you need a single small h
573 better name), loaded by both mptraid a
574 handles registering the /dev/megaraid
575 both mptraid and megaraid would regist
576 adapter discovered, and lsiioctl would
577 redirecting userspace tool ioctls to t
578
579 - Matt Domsch <Matt_Domsch@dell
580
581 ii. Remove C99 initializations from pci_de
582
583 "pci_id_table_g would be much more rea
584 initializers.
585 PCI table doesn't change, there's lots
586 readable variant. And it's really far
587 lines without C99 initializers."
588
589 - Christoph Hellwig <hch@infrad
590
591 iii. Many fixes as suggested by Christoph H<
592 linux-scsi, 05.28.2004
593
594 iv. We now support up to 32 parallel ioctl
595 There is a conscious effort to let mem
596 commands.
597
598 v. Do away with internal memory managemen
599 instead.
600
601 vi. Kill tasklet when unloading the driver
602
603 vii. Do not use "host_lock', driver has fin
604 data structures.
605
606 viii. Optimize the build scatter-gather list
607 know the data transfer address and len
608
609 ix. Better implementation of error handlin
610 performs extended errors recovery for
611
612 x. Disassociate the management commands w
613 Driver now treats the management packe
614 dedicated callback routine.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.