KD64: Raise to HIGH_LEVEL when entering trap
KDBG: lower to DISPATCH_LEVEL when applying IRQL hack & use a worker thread to load symbols
KD&KDBG: Actually unload symbols when required
- Implement caching of individual (newline-separated) text lines; this
behaviour can be enabled with a flag (enabled by MORE):
CON_PAGER_CACHE_INCOMPLETE_LINE.
This feature is necessary when reading a text file, whose text lines
may span across two or more successive temporary read buffers, and is
required for correctly determining whether the lines being read are
blank and may be squeezed.
- When squeezing blank lines, the blank-line check must be done for each
line segment corresponding to the screen line (and following) that
need to be displayed. This matches the behaviour of MS MORE.COM.
- Fix the IsBlankLine() check to not consider FORM-FEEDs as being blank
characters: This is necessary for correctly handling FORM-FEED
expansion. Also note that MS MORE.COM only checks for spaces and TABs,
so we are slightly overdoing these checks (considering other types of
whitespace).
- Get rid of ConCallPagerLine() and the intermediate CON_PAGER_DONT_OUTPUT
state flag that were used repeatedly for each and every small line
chunks. Instead, call directly the user-specified 'PagerLine' callback
when we are about to start treating the next line segment to be
displayed (see comment above).
- Fix the exit return condition of ConPagerWorker(): it should return
TRUE whenever we displayed all the required lines, and FALSE otherwise.
Otherwise, the previous (buggy) condition on the data being read from
the text file, may lead to the prompt not showing when a screenful of
text has been displayed, if it happened that the current text buffer
becomes empty at the same time (even if, overall, the text file hasn't
been fully displayed).
- In MorePagerLine(), when we encounter for the first time a blank line
that will be squeezed with other successive ones, display a single
blank line. But for that, just display one space and a newline: this
single space is especially needed in order to force line wrapping when
the ENABLE_VIRTUAL_TERMINAL_PROCESSING or DISABLE_NEWLINE_AUTO_RETURN
console modes are enabled. Otherwise the cursor remains at the
previous line (without wrapping), and just outputting one newline will
not make it move past 2 lines as one would naively expect.
- The column extent should cover the whole width of the console screenbuffer.
On the contrary, the line extent covers only the number of lines that
cover the current displayed console window.
- Since the console can be resized while the pager is prompting, we need
to actually recalculate in the prompt routine the default number of
lines for a screen.
- Reset s_nNextLineNo when paging a new file.
- Shorten the names of the flags: CON_PAGER_FLAG_*** --> CON_PAGER_***.
- Reorder few members in the CON_PAGER pager structure where they make sense.
- ScreenColumns, ScreenRows --> PageColumns, PageRows: keeping the open
possibility for having a pager controlling a region of different size
than the screen.
- Add doxygen documentation for ConWritePaging().
GCC has some functions, variables & type attributes which can be used as aliases
for some of the SAL annotations. Although it's not as rich & precise, it's still useful
since we actually enable -Werror on GCC builds whereas we don't use such an option
on MSVC builds.
For now, _Must_inspect_result_ is aliased to warn_result_unused attribute.
- Add as well the undocumented 0x40 dwFlags value from shlwapi!PathFileExistsDefExtW()
to the list.
- Fix a bug (also present in Win2k3) in PathResolveW() where, specifying
the PRF_DONTFINDLNK flag, would also erroneously exclude checking for
the .cmd extension as well (which was obviously NOT the original
intention, from the name of the flag and the documentation as well:
https://docs.microsoft.com/en-us/windows/win32/api/shlobj_core/nf-shlobj_core-pathresolve
).
Addendum to 3a822e4f.
Spinlocks are not reentrant (and this is done a lot), using them forces us to have
an horrible hack in the kernel, which unschedules threads which are at DISPATCH_LEVEL
thus allowing another thread to take ownership of the spinlock while the unscheduled
thread should already hold it.
CORE-6473
This speeds up boot a bit and also decreases binary size.
- Introduce a new CMake define ISAPNP_ENABLE that defined only in x86 platform
when ISA PnP is supported.
NEVER DO THIS! It is guaranteed to be wrong. Instead always individually pack single structures that need packing.
This fixes USB mouse on 64 bit builds.
In addition to that, here are some stuff done in this commit whilst testing:
- ICIF_QUERY_SIZE_VARIABLE and friends were badly misused, they should be used only when an information class whose information length size is dyanmic and not fixed. By removing such flags from erroneous classes, this fixes the STATUS_INFO_LENGTH_MISMATCH testcases.
- Use CHAR instead of UCHAR for classes that do not need alignment probing, as every other class in the table do, for the sake of consistency.
- ProcessEnableAlignmentFaultFixup uses BOOLEAN as type size, not CHAR. This fixes a testcase failure on ROS.
- Check for information length size before proceeding further on querying the process' cookie information.
- ProcessHandleTracing wants an alignment of a ULONG, not CHAR.
- Move PROCESS_LDT_INFORMATION and PROCESS_LDT_SIZE outside of NTOS_MODE_USER macro case. This fixes a compilation issue when enabling the alignment probing. My mistake of having them inside NTOS_MODE_USER case, sorry.
- On functions like NtQueryInformationThread and the Process equivalent, complete probing is not done at the beginning of the function, complete probing including if the buffer is writable alongside with datatype misalignment check that is. Instead such check is done on each information class case basis. With that said, we have to explicitly tell DefaultQueryInfoBufferCheck if we want a complete probing or not initially.
KiSetTrapContext is an asm wrapper around RtlSetUnwindContext, which first stores an exception frame to assure that all non-volatile registers were put on the stack, then calls RtlSetUnwindContext to update their first saving positions on the stack and finally restore the exception frame to potentially load any updated registers, that haven't been saved elsewhere on the stack.
RtlpCaptureNonVolatileContextPointers walks the stack and captures the addresses of all non-volatile registers on the stack, when they have been saved first. This is needed to be able to fix up the non-volatile on a system call, which doesn't capture non-volatiles, but relies on them to be restored by the callees.
Instead of only checking for the TargetFrame, also check for a mode change, i.e. RIP went from kernel to user, in which case the target frame was not reached yet, because it was too large, but processing can't continue here.
RtlSetUnwindContext uses RtlpCaptureNonVolatileContextPointers to set the non-volatile registers in the the stack. They will be picked up, when returning back or unwinding, e.g. to the system call handler.
