Hello Hurd,
I made some changes to this patch:
1- I added a call to pci_system_cleanup(); in the shutdown RPC,
so libpciaccess is shutdown correctly.
2- As now we're using libpciaccess functions to read/write
from netfs_attempt_read/write, and these two libpciaccess
functions have diffe
From: Joan Lledó
* pci-arbiter/Makefile:
* Remove pci_access.c and x86_pci.c from the sources.
* pci-arbiter/func_files.c:
* config_blog_op: Call the proper i/o function.
* io_config_file: Use a harcoded PCI config size.
* read_rom_file:
Grab the fu
Hello,
El 27/10/19 a les 16:32, Samuel Thibault ha escrit:> Hello,
> Could you try to split your changes in separate patches?
I splitted the patch into four patches, the first one is the Damien's
original patch adapted to the current head + a changelog.
The other three patches are my changes.
From: Joan Lledó
* pci-arbiter/netfs_impl.c:
* netfs_attempt_write: Cast op function to pci_io_op_t
---
pci-arbiter/netfs_impl.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/pci-arbiter/netfs_impl.c b/pci-arbiter/netfs_impl.c
index 0be8c370..b987a0bc 100644
--- a/p
From: Joan Lledó
* pci-arbiter/startup-ops.c:
* S_startup_dosync: Call pci_system_cleanup().
---
pci-arbiter/startup-ops.c | 4
1 file changed, 4 insertions(+)
diff --git a/pci-arbiter/startup-ops.c b/pci-arbiter/startup-ops.c
index f3506c42..eb387fd9 100644
--- a/pci-arbiter/start
From: Joan Lledó
* pci-arbiter/pcifs.c:
* create_dir_entry:
Limit to NAME_SIZE-1 when calling strncpy().
Finish entry->name with '\0'.
* create_fs_tree:
memset() to 0 the directory entry.
Limit to NAME_SIZE-1 all call
From: Damien Zammit
This patch removes all embedded pciaccess code from the arbiter
and instead uses the external pciaccess library.
* pci-arbiter/Makefile:
* Remove pci_access.c and x86_pci.c from the sources.
* pci-arbiter/func_files.c:
* io_config_file: Use a harcoded PCI conf
From: Joan Lledó
---
src/hurd_pci.c | 83 +-
1 file changed, 34 insertions(+), 49 deletions(-)
diff --git a/src/hurd_pci.c b/src/hurd_pci.c
index 28bef16..98bf83e 100644
--- a/src/hurd_pci.c
+++ b/src/hurd_pci.c
@@ -304,8 +304,8 @@ pci_device_hurd
Hi,
El 3/11/19 a les 17:10, Samuel Thibault ha escrit:
>> +if (!devices)
>> +return ENOMEM;
>
> That makes me realize: we need to closedir(dir).
What do you mean?
>> if (err) {
>> -mach_port_deallocate (mach_task_self (), pci_server_port);
>> -/*
From: Joan Lledó
* pci-arbiter/pcifs.c:
* create_fs_tree:
Remove all memset() previous to snprintf() calls.
---
pci-arbiter/pcifs.c | 4
1 file changed, 4 deletions(-)
diff --git a/pci-arbiter/pcifs.c b/pci-arbiter/pcifs.c
index cc08fad0..aaee7f0b 100644
--- a/pci-arbiter
Hi,
El 3/11/19 a les 21:48, Samuel Thibault ha escrit:
> Hello,
>
> Uploaded fixed netdde, pciutils, libpciaccess, and commited this
> pci-arbiter cleanup!
>
Great!
Now I have some questions:
- You said libpciaccess upstream code for the Hurd doesn't match the one
in the Debian package. Where
El 9/11/19 a les 10:48, Joan Lledó via Bug reports for the GNU Hurd ha
escrit:
> Now I have some questions:
>
> - You said libpciaccess upstream code for the Hurd doesn't match the one
> in the Debian package. Where is the debian repo for the package? I found
> [1] but it
From: Joan Lledó
* pci-arbiter/pcifs.c:
* init_file_system: Remove unnecessary free()
* create_fs_tree:
Fix a boundary overrun where no devices are found.
---
pci-arbiter/pcifs.c | 7 ++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/pci-arbiter/pcifs
From: Joan Lledó
---
src/hurd_pci.c | 153 +
1 file changed, 91 insertions(+), 62 deletions(-)
diff --git a/src/hurd_pci.c b/src/hurd_pci.c
index 28bef16..26505f3 100644
--- a/src/hurd_pci.c
+++ b/src/hurd_pci.c
@@ -304,8 +304,8 @@ pci_device_hurd
Hi,
I'll try to clear things:
1- I wrote this patch from and for 0.16 in upstream:
https://gitlab.freedesktop.org/xorg/lib/libpciaccess
2- Didn't make any PR b/c I was waiting for comments.
3- Didn't thought on Debian at all.
4- If using a weak reference is not good enough for upstream, OK, w
Hi,
What about this? Do I send the PR? Is there any alternative for the
weak reference?
El dt. 26 de 11 de 2019 a les 22:53 +0100, en/na Samuel Thibault va
escriure:
> Joan Lledó, le lun. 25 nov. 2019 10:02:55 +0100, a ecrit:
> > 2- Didn't make any PR b/c I was waiting for comments.
>
> Ok, now
Hello Hurd,
To test nested arbiters, I installed a translator in
/home/jlledom/gnu/servers/bus/pci which connects to the main translator
in /servers/bus/pci. Then, logged as 'demo' user, I use remap to ensure
all rpcs are sent to my translator, not the main one:
demo@debian:~$ remap /servers/bus/
Hi,
El 8/12/19 a les 14:49, Samuel Thibault ha escrit:
> Hello,
>
> Joan Lledó via Bug reports for the GNU Hurd, le dim. 08 déc. 2019 09:04:09
> +0100, a ecrit:
>> demo@debian:~$ remap /servers/bus/pci /home/jlledom/gnu/servers/bus/pci
>> -- rpctrace -o /home/demo/demo.log env
>> LD_LIBRARY
Hi,
El 11/12/19 a les 2:03, Samuel Thibault ha escrit:
> So indeed, when servers/bus/pci/ gets opened, it's not getting remapped.
> Feel free to fix the FIXME :)
Yes, that was it, I wrote a patch (attached) and it worked.
> I would say that pciutils should be made to use stat instead of
> lstat,
From: Joan Lledó
* trans/remap.c:
* trivfs_S_dir_lookup():
* Match and replace prefixes instead of complete
file names. This is needed to remap entire file
systems, not only trivial ones.
---
trans/remap.c | 12
1 file chan
From: Joan Lledó
* trans/remap.c:
* trivfs_S_dir_lookup():
* Match and replace prefixes instead of complete
file names. This is needed to remap entire file
systems, not only trivial ones.
---
trans/remap.c | 25 +
Hello Hurd,
I updated the pciutils patch I wrote in 2017. The new version is attached.
This is the same patch with minor changes:
- Makefile updated to be applied on newer sources.
- hurd_detect(): Use stat() instead of lstat() to work with symbolic links.
- enum_devices(): Call closedir() on err
From: Joan Lledó
A new module for the Hurd that accesses PCI bus using available RPCs.
All references to the Hurd in the i386 access method have been removed.
---
lib/Makefile | 7 +-
lib/configure | 4 +-
lib/hurd.c | 381 +
li
Hi,
El 1/12/19 a les 1:10, Damien Zammit ha escrit:
> libpciaccess 0.16-1+hurd.1 is still broken on rumpdisk. Somehow the 2
> applied patches
> don't allow probing of AHCI to occur in rump_init().
> I've packaged rump libs into a debian package following Robert Millan's work
> and sent it to Sam
Hello Hurd,
I'm glad to say that our patches for both picutils and libpciaccess are
now upstream.
El 10/11/19 a les 2:44, Samuel Thibault ha escrit:
> Joan Lledó, le sam. 09 nov. 2019 17:36:19 +0100, a ecrit:
>> I was also wondering if after your changes, libpciaccess would support
>> nested arbi
From: Joan Lledó
Rename some CLI options in order to add the new -d --device arg.
Replace -s by -c for subclasses.
Replace -D by -d for domains.
Domains are optional from now on, default to 0.
Replace -b by -B for buses.
Bus now creates a new permission scope if the current one
Hello Hurd,
I made some changes in the options parser to add a new --device option, as a
shortcut for current -D -b -d -f options.
I also renamed some options in order to free -D and --device to use it for this
purpose. From now on, the formerly called "devices" are now called "slots", and
"devic
From: Joan Lledó
Shortcut for -d, -B, -s and -f
Usage: --device [:]:.
E.G. --device 00:05.0 is shortcut for -d 0 -B 0 -s 5 -f 0
* pci-arbiter/options.c: Implement --slot option
* pci-arbiter/options.h: Add --slot to options list
---
pci-arbiter/options.c | 60 +++
Hi,
El 15/2/20 a les 15:02, Samuel Thibault ha escrit:
> Why a capital b? It's be more coherent to have -d -b -s -f all
> small-caps.
>
Because of this notice in --help:
"More than one permission scope may be specified. Uppercase options
create a new permission scope if the current one already
From: Joan Lledó
Shortcut for -d, -b, -s and -f
Usage: --device [:]:.
E.G. --device 00:05.0 is shortcut for -d 0 -b 0 -s 5 -f 0
* pci-arbiter/options.c: Implement --slot option
* pci-arbiter/options.h: Add --slot to options list
---
pci-arbiter/options.c | 57 +++
From: Joan Lledó
Rename some CLI options in order to add the new -D --device arg.
Replace -s by -c for subclasses.
Replace -D by -d for domains.
Domains are optional from now on, default to 0.
Replace -d by -s for devices.
The formerly called "devices" are now called "slots", and
Hello,
I'm blocked trying to solve a bug I found in the option parsing of the
arbiter, at:
http://git.savannah.gnu.org/cgit/hurd/hurd.git/tree/pci-arbiter/options.c#n135
strtol() returns 0 if it wasn't able to convert the input, for instance,
-d jj will set h->curset->domain to 0, which means gi
From: Joan Lledó
A misuse of strtol() caused wrong parameters to be interpreted as '0'
* pci-arbiter/options.c:
* New function parse_number() to handle wrong input
* Call parse_number() from all places where strtol was being called
---
pci-arbiter/options.c | 58
Forget about this, it was working with settrans b/c this command replaces the ol
d translator. I was testing the old code when using fsysopts... :(
Anyway, here's the patch to fix that bug.
34 matches
Mail list logo
