- initial support for MAP_AUTOGROW flag. When mapping beyond the end of file is
requested with MAP_AUTOGROW, if pages beyond the end of file are touched, the
file should be resized. We are not able to emulate this yet, so we immediatly
resize the file to fit the whole mapping.
- implements mmap64
address 0x200000 (disasembling usinit shows that this address is hardcoded in
libc). It uses it for locks and semaphres.
We therefore allocate this page of memory, to prevent IRIX process from
faulting when thay call usinit(3).
- the signal trampoline address is given to the kernel by a sigaction()
fourth argument
- we introduce an irix_emuldata structure to keep track of the signal
trampoline address
- we don't support per-sigaction signal trampolines, we only do per-process
- now that we use the IRIX libc signal trampoline, we do not have to handle
the errno update from the signal trampoline
- it is possible that IRIX 5 signal delivery works too, since theses binaries
will come with their own signal trampoline
when SA_SIGINFO is used. The IRIX process will hence find the expected
information using the third argument of the signal handler.
We do not provide code and siginfo yet.
swapent, or as seen in userland, is dbtob(1), which turns to be 512 for all
arch for now.
In struct swapdev, there is another field for block size. This value is private
to uvm_swap.c and is only used for swap I/O on regular files. It is equal to
the underlying device block size and it is not necessarily 512.
- Added two more swapctl commands: GETFREESWAP and GETSWAPVIRT.
There is a problem in the way swap block size are found here. See comment
in get_block_size().
sys_swapctl(SWAP_STATS). This enable the use of a kernel based
buffer instead of using some temporary memory in the stackgap,
whereas we cannot make sure that the size os the struct swapent array
will fit in it. (it is not known at build time, but the stackgap len
is set at build time).
look for a block of free virtual memory big enough to hold all sections. The
blocks starts at the beginning of the first section and ends at the end of
the last section. In the previous version the block ended at the beginning
of the last section, hence creating situations where there was not enough
free space to map the section.
structure for the mountid, but it is 32/64 bits long only, whereas
mountid is 128 bits long. Because we did not initialize the unused bits to
zero, the mountid was not always unique within a filesystem.
This makes autocad 1.3 able to start up.
makes X11 binaries able to actually work: most of them were previously hang
in infinite loop wiaiting for data from the X server because SIOCNREAD
reported that some data where to be read whereas the X server had nothing
to say.
Tested (and works): xlogo, xterm, ghostview (IRIX build). Things are getting
interesting...
header to distinguish between o32, n32 and n64 ABIs. We now use this.
This suppress the need of the mips_option test, which had some fake positive.
This also removes the mandatory ordering of n32 vs o32 in the exec switch
(exec_conf.c)
- do not save/and restore registers that should not be saved and restore
- do give an accurate sigcontext pointer to the signal handler
- do use the struct sigreturna from IRIX.
This eliminates panics and hangs in certain circonstances
Also some cosmetic changes with tabs usage
that the load addresses in the section array are increasing, and that no
section in the array overlap with each other. IRIX proably makes the same
assumptions, but this has not been tested.
The key point with relocation is to always use the same offset for each
section. Because userland gets only the load address off the first section, it
has to assume that all the remaining sections kept the same offset with
respect to the first section. By using fixed offset instead of finding
some free space for each section, we can eliminate the libX11 load hack.
even for program with a DSO using overlapping load virtual addresses.
The fix is a mean hack, see the comments in irix_syssgi.c. It would be nice to
get uvm_map_findspace() to return the page we suggest instead of the page which
is 16384 bytes away.
we used load_psection, then ran each vmcmd and tried to relocate the failing
ones. This fails if there is two vmcmd for one section, and the second is
not a mapping (for instance a map_pagedvn and a map_zero), because the first
one gets relocated, but not the second one.
Additionnaly, it was not necessary to update the userlevel psection array:
libc stubs seems to do the job themselves.
for the text section of libx11.so was overlapping with other ELF sections
aloready loaded, and this resulted into an ENOMEM error.
syssgi(MAPELF) uses elf32_load_psection() from syssrc/sys/kern/exec_subr.c
The problem was never experienced with load_psection() because it only has
to load one section, hence the requested address are not already allocated.
The fix is done when the initial mapping at the default address fails by
finding a free location in the VM space using uvm_map_findspace(), and then
retrying to load the section.
Other details:
- once the ELF section has been relocated, the ELF program header must be
updated with the new address and copied back to userland. For now we always
do it, maybe we could copy it only when it was modified.
- We are able to emulate the exact address where IRIX loads libX11.so instead
of the default location
Service happy. Code in libc attempts to open files in the ns filesystem, and
then uses getmountid on failure to ensure that the ns filesystem is really
mounted. We don't emulate the ns filesystem yet, but getmountid now correctly
reports that ns is not present.
Note: It seems that the mountid of the ns filesystem should always be
00000005 00000000 00000000 7fff3000
manu