- 09 Jan, 2023 40 commits
-
-
Eric Farman authored
It was suggested [1] that we replace the old copy_from_iova() routine (which pins a page, does a memcpy, and unpins the page) with the newer vfio_dma_rw() interface. This has a modest improvement in the overall time spent through the fsm_io_request() path, and simplifies some of the code to boot. [1] https://lore.kernel.org/r/20220706170553.GK693670@nvidia.com/Suggested-by:
Jason Gunthorpe <jgg@nvidia.com> Signed-off-by:
Eric Farman <farman@linux.ibm.com> Reviewed-by:
Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by:
Heiko Carstens <hca@linux.ibm.com>
-
Eric Farman authored
The output of vfio_ccw is always a Format-2 IDAL, but the code that explicitly sets this is buried in cp_init(). In fact the input is often already a Format-2 IDAL, and would be rejected (via the check in ccwchain_calc_length()) if it weren't, so explicitly setting it doesn't do much. Setting it way down here only makes it impossible to make decisions in support of other IDAL formats. Let's move that to where the rest of the ORB is set up, so that the CCW processing in cp_prefetch() is performed according to the contents of the unmodified guest ORB. Signed-off-by:
-
Eric Farman authored
Currently, vfio-ccw copies the ORB from the io_region to the channel_program struct being built. It then adjusts various pieces of that ORB to the values needed to be used by the SSCH issued by vfio-ccw in the host. This includes setting the subchannel key to the default, presumably because Linux doesn't do anything with non-zero storage keys itself. But it seems wrong to convert every I/O to the default key if the guest itself requested a non-zero subchannel (access) key. Any channel program that sets a non-zero key would expect the same key returned in the SCSW of the IRB, not zero, so best to allow that to occur unimpeded. Signed-off-by:
-
Eric Farman authored
There's no need to send in both the address of the subchannel struct, and an element within it, to populate the ORB. Pass the whole pointer and let cp_get_orb() take the pieces that are needed. Suggested-by:
-
Eric Farman authored
There is no longer an mdev struct accessible via a channel program struct, but there are some artifacts remaining that mention it. Clean them up. Signed-off-by:
-
Sven Schnelle authored
Normally a user can scroll back with PF7/PF8 if printed messages are outside of the visible screen area. This doesn't work when the kernel crashes, because the scrollback handling is done by the kernel, which is no longer alive after the kernel crash. Add code to always print all dirty lines in the screen buffer, so the user can scroll back with the terminal scrollback keys (Page Up/Down). Signed-off-by:
Sven Schnelle <svens@linux.ibm.com> Acked-by:
-
Sven Schnelle authored
Now that lines are converted during output, the RA and SBA no longer need to get updated as an additional step. Instead set them when converting the line. Signed-off-by:
-
Sven Schnelle authored
Make TTY3270_UPDATE_ALL the sum of all TTY3270_* flags, so we don't need any special handling for it. Signed-off-by:
-
Sven Schnelle authored
When activating the view fails (in this case because the 3270 is disconnected) return from the notifer callback. Otherwise the system will deadlock. Signed-off-by:
-
Sven Schnelle authored
Use __packed __aligned instead of __attribute__((packed, aligned(X))); to match the rest of the file. Signed-off-by:
-
Sven Schnelle authored
In order to use the fs3270 one would need at least the ioctl definitions in uapi. Add two new include files in uapi, which contain: fs3270: ioctl number declarations + returned struct for TUBGETMOD. raw3270: all the orders, attributes and similar stuff used with 3270 terminals. Signed-off-by:
-
Sven Schnelle authored
fs3270 uses EWRITEA to clear the screen when a user opens /dev/3270/tub. However it misses the attribute byte after the EWRITEA, so (at least) x3270 complains about 'Record too short, missing write flags'. Add the missing flag byte to fix this. Signed-off-by:
-
Sven Schnelle authored
fix function prototypes split over two lines like: static void foobar(void) Signed-off-by:
-
Sven Schnelle authored
Signed-off-by:
-
Sven Schnelle authored
remove a duplicate assignment reported by checkpatch. Signed-off-by:
-
Sven Schnelle authored
Fix a few missing braces and wrong placement of braces reported by checkpatch. Signed-off-by:
-
Sven Schnelle authored
Fix a few whitespace errors reported by checkpatch, namely superfluous whitespace, missing spaces and empty lines. Signed-off-by:
-
Sven Schnelle authored
Fix the following checkpatch warning: CHECK: Comparison to NULL could be written "!rp" + if (rp == NULL) Signed-off-by:
-
Sven Schnelle authored
move the core processing to __raw3270_activate_view() to reduce the required if/else blocks and indentiion levels. Signed-off-by:
-
Sven Schnelle authored
WARN_ON_ONCE if list is not empty, and return an error code instead. Signed-off-by:
-
Sven Schnelle authored
Signed-off-by:
-
Sven Schnelle authored
Add a small comment to the lock member of struct raw3270_view to make checkpatch happy. Signed-off-by:
-
Sven Schnelle authored
checkpatch complains about missing argument names in function declarations. Fix it. Signed-off-by:
-
Sven Schnelle authored
Signed-off-by:
-
Sven Schnelle authored
Fixes a few checkpatch warning about EXPORT_SYMBOL being at the end of the file instead of being next to the functions. Signed-off-by:
-
Sven Schnelle authored
Fix remaining checkpatch issues, like misplaced brackets, whitespace and similar things. No functional change. Signed-off-by:
-
Sven Schnelle authored
Use msecs_to_jiffies() instead of HZ/10. Signed-off-by:
-
Sven Schnelle authored
fix the following and similar checkpatch warnings: CHECK: multiple assignments should be avoided + tp->cx = tp->saved_cx = 0; Signed-off-by:
-
Sven Schnelle authored
fix the following and similar checkpatch warnings: CHECK: Avoid CamelCase: <ESnormal> + enum { ESnormal, ESesc, ESsquare, ESparen, ESgetpars }; Signed-off-by: -
Sven Schnelle authored
To let the user know about function key bindings, print them next to the Running/History field at the lower right of the screen. Also print the scrollback position. Signed-off-by:
Niklas Schnelle <schnelle@linux.ibm.com> Signed-off-by:
-
Sven Schnelle authored
With the previous change this reduces the size of struct tty3270_attribute from four to two bytes. As we have this struct allocated for each character cell, this saves quite some memory. Signed-off-by:
-
Sven Schnelle authored
As we're only supportign 8 colors, we don't need 8 bits. Reduce the size to 4 bits to save memory. Signed-off-by:
-
Sven Schnelle authored
Command line recalling is the last user of the 3270 custom malloc() like allocator. Remove this dependency by using a statically allocated buffer for the saved command lines, and also remove the allocator. Signed-off-by:
-
Sven Schnelle authored
Preparation patch to allow removing the custom 3270 memory allocator. Signed-off-by:
-
Sven Schnelle authored
The length of the screen line is variable with the 3270 protocol. For each attribute (foreground, background color, highlighting etc) we need 3 bytes: the set attribute order, the attribute number, and the value of the attribute. This means that depending on screen content, we might end up 3*3 bytes addtional data for a single character. Allocating the maximum possible amount of memory is quite a lot, and we cannot easily extend the lines by allocating memory because we might get called from atomic context. Failing to extend the memory would also be bad as that would mean that we could miss kernel messages in oom conditions. Therefore move the conversion to a 3270 datastream to tty3270_update(), and use only single line buffer. Signed-off-by:
-
Sven Schnelle authored
Updating the status line is almost the same as generating it when redrawing the screen. However, the code is much easier to read when doing so. Signed-off-by:
-
Sven Schnelle authored
There are lots of places adding attributes or orders to the datastream. Add a few helpers to make that code shorter and easier to read. Signed-off-by:
-
Sven Schnelle authored
Instead of always converting the character set, only convert them when the line is really displayed. Signed-off-by:
-
Sven Schnelle authored
To make the upcoming patches easier to read, move tty3270_convert_line() before changing code. No functional change. Signed-off-by:
-
Sven Schnelle authored
All callers of raw3270_buffer_address() are calculating the offset from some x/y coordinates. Move that calculation inside of the function, so user can pass the x/y values directly. Note that negative values are relative to the end-of-line or end-of-screen. Signed-off-by:
-
