1 QNX编译
1.1 基本概念
QNX可以直接使用Linux Makefile编译库和二进制,在Makefile文件中指定CC=aarch64-unknown-nto-qnx7.0.0-g++,或者CC=x86_64-pc-nto-qnx7.0.0-g++,保存退出后,运行source /qnx_sdk_path/qnxsdp-env.sh,然后再运行make即可。
QNX官方不建议直接使用xxx-g++,而是使用q++ -Vxxx,q++通过选项-Vxxx找到对应的g++编译器,如何获取xxx,可以使用q++ -V查询。并且只能将选项-Vxxx加到C语言的CFLAGS或者C++语言的CXXFLAGS中,不能加到其它地方,否则编译会出现各种奇怪的错误。
在q++编译中宏-D_QNX_SOURCE的作用是包括POSIX的头文件,包括数据类型、宏声明、库函数等。如果CXXFLAGS只加了-std=C++11,但没有-D_QNX_SOURCE,就算是包含了头文件unistd.h等,还是无法使用POSIX相关库函数和声明等。C++11先前被称作C++0x。
需要注意的是QNX下Makefile链接线程库使用-pthread而不是Linux风格的-lpthread,而其它的库仍然和Linux一样,需要添加前缀-l,例如-lcrypto。
source /qnx_sdk_path/qnxsdp-env.sh
x86_64-pc-nto-qnx7.0.0-gcc \
-o tcp_server tcp_server.c -lsocket
1.2 showcase
CC = q++
.PHONY: clean
OBJS = src/main.o
CFLAGS = \
-I./include \
-Vgcc_ntox86_64_cxx
LDFLAGS = -pthread -lcrypto
compile every cpp file to object
%.o: %.cpp
$(CC) $(CFLAGS) -c $^ -o $@
link all of the objects to hello
hello: $(OBJS)
$(CC) $(CFLAGS) $(OBJS) $(LDFLAGS) \
-o $@
clean:
rm -rf $(OBJS)
rm -rf hello
1.3 QNX proprietary Makefile
variant:
g-le: A debugging version of a little-endian executable
o-le: executable
a-le: static library, a means archive
so-le: shared library
mkdir my_project
cd my_project
mkdir inc
mkdir src
touch common.mk
touch Makefile
addvariant aarch64 o-le
le: little endian
PINFO: package info
1.4 build_files
xxx.build.tmpl:一般一个分区一个build文件。Because .build file does not support MACRO, could use Linux filepp command to replace MACRO.
filepp -DMACRO_TEST=${MACRO_TEST} \
my.build > my.build_new
on -R0x7 pipe &
-R:表示CPU掩码,亲和性;0x7 = 0111,CPU0、CPU1、CPU2,不包含CPU3
mkifs xxx.build.tmpl xxx.ifs
dumpifs xxx.ifs
/proc/boot
2 进程和sepol
2.1 进程优先级
QNX的线程优先级,是一个0 - 255的数字,数字越大优先级越高。 所以,优先级0是内核中的idle线程;同时,优先级64是一个分界岭,就是说,优先级1 - 63是非特权优先级,一般用户都可以用,而64 - 255必须是有root权限的线程才可以设。QNX进程或者线程刚创建时的优先级是10。
procnto-smp-instr: Instrumented
2.2 QNX secpol
secpol是QNX 7.0才引入的,类似于Android sepolicy,运行on命令时指定参数-T xxx_service_t即可。
编译时需要将SDK 7.0中的aarch64le/bin/secpol和aarch64le/bin/secpolmonitor(不能与secpolgenerate同时运行,因为这2个命令用到同一个目录,运行有冲突)集成到image中。
allow_attach: resmgr_attach()
allow_link: pathmgr_symlink()
SECURE_BOOT_SECPOL
allow_attach io_pkt_v6_hc_t {
/path/to/socket_name
};
/proc/boot/secpol.bin
secpolmonitor -pasno /tmp/secpol.out &
secpolmonitor -pasno /tmp/secpol.out \
-S subrangedAble &
genpol
2.3 列出所有的进程加载了哪些so库
pidin -F "%O"
3 console
3.1 Android vdev console
x86 vdev ser8250
aarch64 ttyAMA0
fdtdump, qnx Configuring guests
vdev vdev-pl011.so
loc 0x1c090000
intr gic:37
3.2 命令
To attach to an existing session:
dtach -a <socket> <options>
Detach from a session
In an attached session, type Ctrl+\
QVM <-> dtach server <-> dtach client <-> QNX shell
ptsname slave (QNX QVM) <-> /dev/ptmx master (QNX dtach, other node by QNX) <-> /tmp/console (QNX dtach, unix-domain socket server) <-> dtach client (dtach -a /tmp/console) <-> QNX shell
3.3 dtach showcase
server:
on -T xxx_t -u 0:0 \
-d /usr/bin/dtach \
-n /path/to/socket_name \
-r winch qvm @/vm/images/linux-la.config
-T apply secpol type to dtach
-u uid:gid
-d daemon
-n new session
-r redraw, winch means windows change
client:
dtach -a /path/to/socket_name
-a attach
3.4 slogger
https://github.com/christianzeroc/openqnx
slog2info是slogger2的客户端。
4 Block
4.1 Block设备
/dev/hd0t12 - t表示type,12表示FAT32分区
/dev/hd1t6 - t表示type,6表示文件系统类型FAT
/fs/usb0
fdisk /dev/hd1 show
dinit:disk initialization
[25-Jan-2022]
mount block device as read write, find command from .build file.
mount -tqnx6 -u -w /dev/xxx /xxx
4.2 Android和QNX共享数据
QNX ->
ls /dev/hd0.*
mkdosfs /dev/hd0.xxx
mount -t dos /dev/hd0.xxx /mnt
Android ->
ls -al /dev/block/by-name/
mount -t vfat -o sync,rw /dev/block/vda17 /mnt/share
5 Display
5.1 keyboard
vdev vdev-virtio-input.so
/usr/share/keyboard/US_101.kbd
slog2info | grep screen
slog2info | grep qvm
5.2 screen
#include <screen/screen.h>
screen_set_event_property_iv()
iv: integer value
cv: char value
6 Network
6.1 FreeBSD
QNX network commands all come from BSD family.
ral0: devnp-ral.so, Ralink
wm0: devnp-e1000.so, wiseman, name comes from FreeBSD
nw_dll_syms
https://github.com/ownmac/qnx_drv/tree/master/hardware/devnp/e1000
6.2 Basic Commands
io-pkt-v6-hc: hardware crypto, crypto is OpenSSL library.
TUNnel: peer=/dev/qvm/la/la_to_host,\
bind=/dev/vdevpeer/vp0,
on io-pkt-v6-hc -d vdevpeer-net \
peer=/dev/qvm/la/la_to_host,\
bind=/dev/vdevpeer/vp0,\
mac=bbbbbbbbbbbb
mount -T io-pkt -o <options>
options for network driver
unload eth0 driver
call driver xxx_detach()
ifconfig eth0 destroy
ifconfig bridge0 create
ifconfig bridge0 destroy
brconfig bridge0 delete eth0
6.3 iptables NAT for hypervisor
PREROUTING:DNAT专用。
POSTROUTING:SNAT专用。
MASQUERADE:SNAT转换时,如果--to-source的地址经常变化,那么就无法在iptables命令中固定,譬如路由器WAN口上的IP地址,内网设备通过WAN口访问外网时,只要指定使用MASQUERADE参数,iptables就会统一将内网设备的IP地址替换为路由器WAN口的IP地址。
Figure 6-1 NAT chain
vp0: 192.168.0.98 (QNX veth)
eth0: 192.168.0.99 (Android veth)
eth1: 192.168.5.99 (Android USB Ethernet)
Because PC could not get QNX vp0 MAC, so use the following commands for NAT, PC client should connect eth1 IP.
client -> eth1 -> eth0 -> vp0 -> server
iptables -I FORWARD 1 -i eth1 -o eth0 -j ACCEPT
iptables -I FORWARD 1 -i eth0 -o eth1 -j ACCEPT
iptables -t nat -A PREROUTING -i eth1 -j DNAT --to-destination 192.168.0.98
iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE
echo 1 > /proc/sys/net/ipv4/ip_forward
6.4 QNX sshd
passwordless登陆,QNX Neutrino 6.6.0用户密码加密默认算法是SHA-512,1000轮计算,并附带16字节的盐值。
-
Windows puttygen产生公匙和私匙.ppk
-
Windows WinSCP导入私匙.ppk
-
将Windows puttygen产生的公匙转换为QNX的格式
QNX shell:
ssh-keygen -i -f win_rsa.pub > /path/to/authorized_keys
- /path/to/sshd_config
StrictModes no
PasswordAuthentication yes
PubkeyAuthentication yes
AuthorizedKeysFile /path/to/authorized_keys
6.5 Windows as ssh server
- Install ssh server for Win10: BvSshServer (original name is WinSSHD)
https://www.bitvise.com/ssh-server-download
Bitvise SSH Server installer
a) Personal Edition
b) Open easy settings
c) Server settings
d) Virtual accounts
-
Virtual account password
-
test/888888
-
Save changes
e) Start Server
- 在QNX环境下,将Windows下的文件复制到QNX系统中:
scp -S /path/to/ssh virtual_account_name@192.168.3.181:/d:/test/config.ips /tmp
如果ssh不在/usr/bin/ssh下,需要给scp指定-S参数,否则scp执行失败。
6.6 QNX Momentics Debug
- QNX shell
Run qconn, default qconn port: 8000
Enable bridge or NAT
- Windows Momentics IDE
[Windows][Perspective][Open Perspective][QNX System Information]
[Target Navigator][New QNX Target]
Configure IP and 8000, then Momentics will connect to QNX automatically.
6.7 eAVB
ptpd-avb: gPTP
ptpd: IEEE1588 v2
6.8 QNX adb test
- Android
ip route get <X.Y.Z.5>
ip rule add to X.Y.Z.0/24 lookup main pref 9999
setprop service.adb.tcp.port 5555
stop adbd
start adbd
iptables-save | grep 5555
iptables -A INPUT -i eth0 -p tcp -m tcp --dport 5555 -j ACCEPT
iptables -A OUTPUT -o eth0 -p tcp -m tcp --dport 5555 -j ACCEPT
- QNX
mount block device as read write, find command from .build file.
mount -tqnx6 -u -w /dev/xxx /xxx
Copy libnbutil.so.1 to /path/to/lib as LD_LIBRARY_PATH shown.
route get <ANDROID_IP>
adb connect <ANDROID_IP>
adb shell
7 x86 USB
7.1 FreeBSD
QNX USB API and umass (src/sys/dev/usb/umass.c) all come from BSD family.
FreeBSD SCSI software architecture is called CAM (Common Access Method), it developed by Future Domain and other SCSI vendors. Linux has little support for a SCSI CAM system yet (mainly for booting from hard disk). CAM even supports target mode, so one could disguise ones computer as a peripheral hardware device (e.g. for a small SCSI net).
7.2 usblauncher_otg
client communicates with io-usb through /dev/io-usb/io-usb
pps: Persistent Publish/Subscribe
/pps/qnx/device/usb_ctrl
usbd_alloc(): /memory/below4G, pidin syspage=asinfo (Address Space Information)
pidin -p qvm irqs
while true; do /bin/slay qvm; \
if [ $? -ne 1 ]; then break; fi; \
done
io-usb-otg -d xhci
devb-umass cam pnp
usb -vvvv
slay devb-umass
7.3 USB PHY Tunning
- QNX设备地址映射函数
#include <sys/mman.h>
#include <sys/neutrino.h>
ThreadCtl(_NTO_TCTL_IO, 0);
void *ptr = mmap_device_memory(0,
len,
PROT_READ|PROT_WRITE| \
PROT_NOCACHE,
0,
0xb8000);
if (ptr == MAP_FAILED) {}
- in32_x86 and out32
get xHCI BAR0 in QNX, skip bit0-bit3 of BAR0
pci-tool -d 0:21:0 --read=CFG:0x10
APL (Gen9, A39X0) 8-port MPH xHCI PORTSC1
in32_x86 <addr from the last step + 0x480>
7.4 x86 xHCI透传
APL (Gen9, A39X0) 8-port MPH xHCI透传给Android后,尽管在QNX下可以使用in32_x86命令和函数mmap_device_memory()读写xHCI寄存器,但是由于IOMMU中仍然有xHCI的DMAR和INTR配置,所以即使卸载Android端xHCI驱动,加载QNX端xHCI驱动,也不会工作;要想让xHCI在QNX下可以工作,需要删除xHCI在IOMMU中的DMAR和INTR配置。
8 x86 GPIO
8.1 P2SB
PCH(Platform Controller Hub)上大部分设备可以通过PCIe或IO方式访问,但PCH上部分设备需要访问PCH的私有空间,这部分空间通过P2SB(Primary to SideBand)的SBREG_BAR寄存器映射到内存空间,这段空间被称为PCR(PCH Private Configuration Space Register)。每个设备对应一个PortID,PortID表示设备在PCR空间的偏移量,再加上寄存器偏移就可以获取寄存器的地址。
8.2 P2SB GPIO
x86下GPIO寄存器位于PCH的私有空间。GPIO被分组,每组对应一个PCR的PortID。GPIO community和PortID的对应关系如下所示。
SouthWest: 0xC0
NorthWest: 0xC4
North: 0xC5
West: 0xC7
每个community的每个GPIO有2个寄存器,分别是PADCFG0(偏移0)和PADCFG1(偏移4),2个寄存器占用8个字节的地址,所以计算某个GPIO 2个寄存器的绝对地址方法如下,其中PortID是每个GPIO community在PCR空间的偏移量。
基地址pad_regs = 0xf8000000(P2SB bar0) + (PortID << 16) + 0x500(read from PADBAR);
某个GPIO padcfg0的绝对地址 = pad_regs + PADCFG0 + 在bank内的GPIO编号 x 8;
某个GPIO padcfg1的绝对地址 = pad_regs + PADCFG1 + 在bank内的GPIO编号 x 8;
每个community的32个GPIO公用2个寄存器,分别是GPI_IS(GPI Interrupt Status)和GPI_IE(GPI Interrupt Enable),2个中断相关寄存器的绝对地址计算方法如下,其中PortID是每个GPIO community在PCR空间的偏移量。
基地址regs = 0xf8000000(P2SB bar0) + (PortID << 16);
某个GPIO GPI_IS的绝对地址 = regs + 0x100(GPI_IS)+ 在bank内的GPIO编号 / 32;
某个GPIO GPI_IE的绝对地址 = regs + 0x110(GPI_IE)+ 在bank内的GPIO编号 / 32;
/sys/class/gpio/gpiochip434/label
跟dts或者ACPI dsl文件对应上,是哪一组
gpiochip434 - North
gpiochip357 - Northwest
gpiochip310 - West
gpiochip267 - Southwest
9 Abbreviations
APIC:读作ei pic
APS:adaptive partitioning
CAM:UNIX/QNX SCSI Common Access Methods
dtach:detach
hogs:List the processes that are hogging the CPU
IFS:Image File System,QNX系统镜像
IPL:Image Program Loader,bootloader程序
mbuf: Unix TCP/IP message buffer
OCP:Open Core Protocol
OpenUSBDI: Open USB Driver Interface
P2SB:Primary to SideBand
pbuf: lwip packet buffer
pidin:类似于Linux ps,也可以执行pidin arg
PPS: QNX Persistent Pusblish/SubScribe
QNX: Quick UNIX
slay <pid>:kill进程
slm: System Launch and Monitor
smmuman:IOMMU/SMMU Manager
ttyAMA: ARM AMBA PL011
vp0: vdevpeer