* Revert "Don't ResizeCache on sector change if no filename is defined (#1438)"
This reverts commit dd9a3296d4.
* Revert "Add ModeSense page 0x25 (DEC special function control page) (#1412)"
This reverts commit 1121b8d9d6.
* Revert "DiskCache needs a size"
This reverts commit 7cc8df271c.
* Revert "Honor sector size change via ModeSelect6 in scsicd (#1406)"
This reverts commit b7f65d33e2.
* Revert "Multiple fixes for ModeSelect (#1405)"
This reverts commit ad5eae93e7.
* Add tests for issue#1426
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* add test without medium
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Don't resize DiskCache without a file
DiskCache computes its size from a file.
Skip ResizeCache Without file (i.e. 'empty' CD)
Fixes#1426
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Add ModeSense page 0x25 (DECSpecialFunctionControlPage)
VAXServer 3100 (CPU KA41-E) console firmware issues ModeSense(6)
page 0x25 when probing for disks. A disk won't be recognized if
this returns an error.
The DEC SCSI specification[1], section 8.5 documents this page.
[1] https://manx-docs.org/collections/antonio/dec/dec-scsi.pdfFixes#1410
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Make ModeSelect() non-const
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Implement ModeSelect for scsicd
and honor sector size setting.
DEC's VMS can't handle 2k sector sizes and uses ModeSelect6 to set the
sector size to 512 bytes.
Fixes#1397
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Test sector size setting via ModeSelect in SCSICD
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Re-calculate total blocks when sector size changes
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Reset CD data tracks after sector size change
The track calculation is based on sector size and must be reset after
change of sector size.
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* resize cache after change of sector size
The disk cache is based on sector size and must be resized when the
sector size changes.
Disk::ResizeCache needs a `raw` parameter, make this value accessible
from the current cache.
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Make GetRawMode const
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
---------
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Allow 'empty' ModeSelect6
tl;dr
Treat a computed length of 0 as `has_valid_page_code`.
Details:
The SRM console (aka 'BIOS') of DEC Alpha sends an empty
ModeSelect6 with the following data:
~~~
ModeSelect6, CDB $151000000c00
~~~
That makes 12 byte(s) as follows
~~~
0 1 2 3 4 5 6 7 8 9 10 11
00 00 00 08 00 00 00 00 00 00 02 00
~~~
decoding it (accoring to [1], Section 8.3.3, Table 94) gives us
Mode Data Length 0
Medium Type 0
Device-specific 0
Block desc len 8
Density Code 0
Number of blks 0
Reserved 0
Block length 512
`scsi_command_util::ModeSelect` computes
~~~
offset = 4 + buf[3];
~~~
giving 12 and
~~~
length -= offset;
~~~
giving 0.
Thus it never enters the `while` loop and `has_valid_page_code` stays
`false`, raising an error.
[1] [Small Computer System Interface - 2 rev 10L.pdf](https://dn790004.ca.archive.org/0/items/SCSISpecificationDocumentsSCSIDocuments/Small%20Computer%20System%20Interface%20-%202%20rev%2010L.pdf)
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Allow ModeSelect with page code 1
OpenVMS Alpha (the operating system, not the SRM BIOS) uses
ModeSelect6 with a page code of 1.
The semantics are unknown, just accepting it works for me.
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Fix page length computation in ModeSelect
tl;dr
The 'skip to next ModeSelect page' computation was off-by-one, either not
taking the page code itself into account or missing the fact that the
page length is given as `n - 1`.
Fix:
Add 1 to the computed length.
Details:
OpenVMS Alpha sends a ModeSelect6 as follows
~~~
command:
ModeSelect6, CDB $151000001900
payload:
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
00 00 00 08 00 00 00 00 00 00 02 00 01 0a 24 00 00 00 00 00 00 00 00 00 00
~~~
This translates to (accoring to [1], Section 8.3.3)
~~~
Mode Data Length 0
Medium Type 0
Device-specific 0
Block desc len 8
~~~
with the following offset / length computation _before_ the `while` loop
~~~
offset = 12
length = 13
~~~
The first payload section is
~~~
4 5 6 7 8 9 10 11
00 00 00 00 00 00 02 00
~~~
translating to
~~~
Density Code 0
Number of blks 0
Reserved 0
Block length 0x200 512
~~~
Then follows a pagecode 1 as
~~~
12 13 14 15 16 17 18 19 20 21 22 23 24
01 0a 24 00 00 00 00 00 00 00 00 00 00
~~~
translating to
~~~~
Page code 1
Page length -1 10
Mode parameters 24 00 00 00 00 00 00 00 00 00 00
~~~
computing (inside the `while` loop, as `// Advance to the next page`)
~~~
size = 10 + 2 = 12
~~~
followed by new `offset` and `length` values
~~~
offset = 25
length = 1
~~~
So it stays in the `while` loop (and has a larger-than-buffer `offset`
value)
Signed-off-by: Klaus Kämpf <kkaempf@gmail.com>
* Improve how commands are internally executed
* Use const CommandContext for execution
* Update error handling
* Fix SonarQube issues
* Remove duplicate code
* Use mutex instead of atomic_bool
* Add hasher
* Add param_map
* Do not log unknown operations as an error for backward/foward compatibility
* Rebrand project to PiSCSI
- rascsi ->piscsi
- rasctl -> scsictl
- rasdump -> scsidump
- ras* -> piscsi* (rasutil -> piscsi_util, etc.)
* Refined the formatting and wording of the app startup banner
* Kept some references to rascsi and rasctl where backwards compatibility is concerned
* Point to the new github repo URL
Co-authored-by: nucleogenic <nr@nucleogenic.com>
Co-authored-by: Uwe Seimet <Uwe.Seimet@seimet.de>
* Support for ID and LUN parameter for the -L option in rascsi
* Added DeviceLogger class
* Removed dupiicate code
* Fixed SonarQube issues
* Added unit tests, improved code sharing
* Fixed regression (#979)
* Cleaned up dependencies on controller manager
* Removed global fields
* Simplified setting up RascsiResponse and RascsiExecutor
* Got rid of remaining raw pointers
* Use references instead of pointers
* Improved encapsulation
* Moved rascsi/rasctl/scsimon/rasdump.cpp to classes (for better testability)
* Moved bus.* to hal folder
* Removed some global variables
* Fixed code redundancies
* Using lambdas instead of member function pointers simplifies the command dispatching and reduces the code volume
* Removed duplicate error handling
* Fix for issue #956
* Unit test updates
* Resolved SonarQube issues
* Unit test updates
* Lambda syntax cleanup
* Use new-style casts
* Use std::none_of when saving the cache
* Use to_integer instead of casts
* Use accessors for getting CDB data
* Made ctrl_t private
* Improved encapsulation
* Replaced pointers by references
* Removed all remaining occurrences of DWORD and BYTE, making os.h obsolete
- Moved C++ code to cpp/ from src/raspberrypi
- Related updates to Makefile, easyinstall.sh, and the github build rules
- Removed the native X68k C code in src/x68k from the repo
