GNU Linux-libre 5.10.219-gnu1
[releases.git] / Documentation / scsi / ChangeLog.arcmsr
1 **************************************************************************
2 ** History
3 **
4 **   REV#         DATE             NAME         DESCRIPTION
5 ** 1.00.00.00    3/31/2004       Erich Chen     First release
6 ** 1.10.00.04    7/28/2004       Erich Chen     modify for ioctl
7 ** 1.10.00.06    8/28/2004       Erich Chen     modify for 2.6.x
8 ** 1.10.00.08    9/28/2004       Erich Chen     modify for x86_64
9 ** 1.10.00.10   10/10/2004       Erich Chen     bug fix for SMP & ioctl
10 ** 1.20.00.00   11/29/2004       Erich Chen     bug fix with arcmsr_bus_reset when PHY error
11 ** 1.20.00.02   12/09/2004       Erich Chen     bug fix with over 2T bytes RAID Volume
12 ** 1.20.00.04    1/09/2005       Erich Chen     fits for Debian linux kernel version 2.2.xx
13 ** 1.20.00.05    2/20/2005       Erich Chen     cleanly as look like a Linux driver at 2.6.x
14 **                                              thanks for peoples kindness comment
15 **                                              Kornel Wieliczek
16 **                                              Christoph Hellwig
17 **                                              Adrian Bunk
18 **                                              Andrew Morton
19 **                                              Christoph Hellwig
20 **                                              James Bottomley
21 **                                              Arjan van de Ven
22 ** 1.20.00.06    3/12/2005       Erich Chen     fix with arcmsr_pci_unmap_dma "unsigned long" cast,
23 **                                              modify PCCB POOL allocated by "dma_alloc_coherent"
24 **                                              (Kornel Wieliczek's comment)
25 ** 1.20.00.07    3/23/2005       Erich Chen     bug fix with arcmsr_scsi_host_template_init
26 **                                              occur segmentation fault,
27 **                                              if RAID adapter does not on PCI slot
28 **                                              and modprobe/rmmod this driver twice.
29 **                                              bug fix enormous stack usage (Adrian Bunk's comment)
30 ** 1.20.00.08    6/23/2005       Erich Chen     bug fix with abort command,
31 **                                              in case of heavy loading when sata cable
32 **                                              working on low quality connection
33 ** 1.20.00.09    9/12/2005       Erich Chen     bug fix with abort command handling, firmware version check
34 **                                              and firmware update notify for hardware bug fix
35 ** 1.20.00.10    9/23/2005       Erich Chen     enhance sysfs function for change driver's max tag Q number.
36 **                                              add DMA_64BIT_MASK for backward compatible with all 2.6.x
37 **                                              add some useful message for abort command
38 **                                              add ioctl code 'ARCMSR_IOCTL_FLUSH_ADAPTER_CACHE'
39 **                                              customer can send this command for sync raid volume data
40 ** 1.20.00.11    9/29/2005       Erich Chen     by comment of Arjan van de Ven fix incorrect msleep redefine
41 **                                              cast off sizeof(dma_addr_t) condition for 64bit pci_set_dma_mask
42 ** 1.20.00.12    9/30/2005       Erich Chen     bug fix with 64bit platform's ccbs using if over 4G system memory
43 **                                              change 64bit pci_set_consistent_dma_mask into 32bit
44 **                                              increcct adapter count if adapter initialize fail.
45 **                                              miss edit at arcmsr_build_ccb....
46 **                                              psge += sizeof(struct _SG64ENTRY *) =>
47 **                                              psge += sizeof(struct _SG64ENTRY)
48 **                                              64 bits sg entry would be incorrectly calculated
49 **                                              thanks Kornel Wieliczek give me kindly notify
50 **                                              and detail description
51 ** 1.20.00.13   11/15/2005       Erich Chen     scheduling pending ccb with FIFO
52 **                                              change the architecture of arcmsr command queue list
53 **                                              for linux standard list
54 **                                              enable usage of pci message signal interrupt
55 **                                              follow Randy.Danlup kindness suggestion cleanup this code
56 ** 1.20.00.14   05/02/2007       Erich Chen & Nick Cheng
57 **                                              1.implement PCI-Express error recovery function and AER capability
58 **                                              2.implement the selection of ARCMSR_MAX_XFER_SECTORS_B=4096
59 **                                              if firmware version is newer than 1.42
60 **                                              3.modify arcmsr_iop_reset to improve the ability
61 **                                              4.modify the ISR, arcmsr_interrupt routine,to prevent the
62 **                                              inconsistency with sg_mod driver if application directly calls
63 **                                              the arcmsr driver w/o passing through scsi mid layer
64 **                                              specially thanks to Yanmin Zhang's openhanded help about AER
65 ** 1.20.00.15   08/30/2007       Erich Chen & Nick Cheng
66 **                                              1. support ARC1200/1201/1202 SATA RAID adapter, which is named
67 **                                              ACB_ADAPTER_TYPE_B
68 **                                              2. modify the arcmsr_pci_slot_reset function
69 **                                              3. modify the arcmsr_pci_ers_disconnect_forepart function
70 **                                              4. modify the arcmsr_pci_ers_need_reset_forepart function
71 ** 1.20.00.15   09/27/2007       Erich Chen & Nick Cheng
72 **                                              1. add arcmsr_enable_eoi_mode() on adapter Type B
73 **                                              2. add readl(reg->iop2drv_doorbell_reg) in arcmsr_handle_hbb_isr()
74 **                                              in case of the doorbell interrupt clearance is cached
75 ** 1.20.00.15   10/01/2007       Erich Chen & Nick Cheng
76 **                                              1. modify acb->devstate[i][j]
77 **                                              as ARECA_RAID_GOOD instead of
78 **                                              ARECA_RAID_GONE in arcmsr_alloc_ccb_pool
79 ** 1.20.00.15   11/06/2007       Erich Chen & Nick Cheng
80 **                                              1. add conditional declaration for
81 **                                              arcmsr_pci_error_detected() and
82 **                                              arcmsr_pci_slot_reset
83 ** 1.20.00.15   11/23/2007       Erich Chen & Nick Cheng
84 **                                              1.check if the sg list member number
85 **                                              exceeds arcmsr default limit in arcmsr_build_ccb()
86 **                                              2.change the returned value type of arcmsr_build_ccb()
87 **                                              from "void" to "int"
88 **                                              3.add the conditional check if arcmsr_build_ccb()
89 **                                              returns FAILED
90 ** 1.20.00.15   12/04/2007       Erich Chen & Nick Cheng
91 **                                              1. modify arcmsr_drain_donequeue() to ignore unknown
92 **                                              command and let kernel process command timeout.
93 **                                              This could handle IO request violating max. segments
94 **                                              while Linux XFS over DM-CRYPT.
95 **                                              Thanks to Milan Broz's comments <mbroz@redhat.com>
96 ** 1.20.00.15   12/24/2007       Erich Chen & Nick Cheng
97 **                                              1.fix the portability problems
98 **                                              2.fix type B where we should _not_ iounmap() acb->pmu;
99 **                                              it's not ioremapped.
100 **                                              3.add return -ENOMEM if ioremap() fails
101 **                                              4.transfer IS_SG64_ADDR w/ cpu_to_le32()
102 **                                              in arcmsr_build_ccb
103 **                                              5. modify acb->devstate[i][j] as ARECA_RAID_GONE instead of
104 **                                              ARECA_RAID_GOOD in arcmsr_alloc_ccb_pool()
105 **                                              6.fix arcmsr_cdb->Context as (unsigned long)arcmsr_cdb
106 **                                              7.add the checking state of
107 **                                              (outbound_intstatus & ARCMSR_MU_OUTBOUND_HANDLE_INT) == 0
108 **                                              in arcmsr_handle_hba_isr
109 **                                              8.replace pci_alloc_consistent()/pci_free_consistent() with kmalloc()/kfree() in arcmsr_iop_message_xfer()
110 **                                              9. fix the release of dma memory for type B in arcmsr_free_ccb_pool()
111 **                                              10.fix the arcmsr_polling_hbb_ccbdone()
112 ** 1.20.00.15   02/27/2008      Erich Chen & Nick Cheng
113 **                                              1.arcmsr_iop_message_xfer() is called from atomic context under the
114 **                                              queuecommand scsi_host_template handler. James Bottomley pointed out
115 **                                              that the current GFP_KERNEL|GFP_DMA flags are wrong: firstly we are in
116 **                                              atomic context, secondly this memory is not used for DMA.
117 **                                              Also removed some unneeded casts. Thanks to Daniel Drake <dsd@gentoo.org>
118 **************************************************************************