全志在线开发者论坛

laocai

我知道的，只有v3s和f1c100主线有drm驱动，而且都是支持不完整的。你可以到挖坑网上去搜索一下，那上面信息多一些。

laocai

@eestr 楼主的项目有进展了吗？我也有兴趣做一个长条的usb屏，可以一起做。

laocai

@yteraa 我补充一点，Tina默认是用的procd，如果要切换到busybox-init, Base system里面要确保procd package被关闭。否则编译的时候会出现如下错误：

Configuring canutils-candump.
Collected errors:
 * check_data_file_clashes: Package busybox wants to install file /opt/tina5sdk-bsp/out/t113_i/maple/openwrt/build_dir/target/root-t113_i-maple/sbin/init
	But that file is already provided by package  * procd
 * opkg_install_cmd: Cannot install package busybox.
package/Makefile:66: recipe for target 'package/install' failed
make[2]: *** [package/install] Error 255
make[2]: Leaving directory '/opt/tina5sdk-bsp/openwrt/openwrt'
package/Makefile:111: recipe for target '/opt/tina5sdk-bsp/openwrt/openwrt/staging_dir/target/stamp/.package_install' failed
make[1]: *** [/opt/tina5sdk-bsp/openwrt/openwrt/staging_dir/target/stamp/.package_install] Error 2
make[1]: Leaving directory '/opt/tina5sdk-bsp/openwrt/openwrt'
/opt/tina5sdk-bsp/openwrt/openwrt/include/toplevel.mk:238: recipe for target 'world' failed
make: *** [world] Error 2
make: Leaving directory '/opt/tina5sdk-bsp/openwrt/openwrt'
INFO: build_openwrt_rootfs failed

laocai

@laocai 为什么今天发帖，都是显示forbidden?

laocai

@zxdplay 谢谢回复，我后来就是这么做的，形成了一个自己专有的头文件。

laocai

现在有两个图层，它们由其他app生成，我希望通过编程的方法，简单的进行切换：显示图层1-->显示图层2-->再显示图层1-- >以此类推。如果通过DISP_LAYER_SET_CONFIG调用的话，似乎要把图层所有的参数，包括缓冲区地址等统统设置一遍，而我只想简单的切换两个图层，并不知道也不关心这些图层的详细参数。请问我该怎么做？

# cat /sys/class/disp/disp/attr/sys | more
screen 0:
de_rate 300000000 hz, ref_fps:62
mgr0: 480x480 fmt[rgb] cs[0x204] range[full] eotf[0x4] bits[8bits] err[0] force_
sync[0] unblank direct_show[false] iommu[1]
dmabuf: cache[0] cache max[0] umap skip[0] umap skip max[18]
        lcd output      backlight(229)  fps:62.1         480x 480
        err:11428       skip:90 irq:14361       vsync:0 vsync_skip:0
   BUF    enable ch[0] lyr[0] z[0] prem[N] a[globl 255] fmt[ 11] fb[ 480, 480;
 0,   0;   0,   0] crop[   0,   0,   0,   0] frame[   0,   0, 480, 480] addr[  2
80000,       0,       0] flags[0x       0] trd[0,0]
depth[ 0]    BUF    enable ch[1] lyr[0] z[16] prem[N] a[pixel 255] fmt[  0] fb[
480, 480; 480, 480; 480, 480] crop[   0,   0, 480, 480] frame[   0,   0, 480, 48
0] addr[       0,       0,       0] flags[0x       0] trd[0,0]
depth[ 0]

laocai

@ddrddr 可能你的framebuffer支持alpha通道，但是填充图像数据的时候没有填写相应alpha数值。检查一下framebuffer格式即可。

laocai

@laocai 官方文档里也到处都是错误。

红线处正确的代码应该是这样的：

        memset(&config, 0, sizeof(struct disp_layer_config));
        config.channel = 0;//blending channel
        config.layer_id = 0;//layer index in the blending channel
        config.enable = 1;
        config.info.mode = LAYER_MODE_BUFFER;

laocai

@zheng1 在关于sunxi图层设置DISP_LAYER_SET_CONFIG的一个问题中说：

@laocai tina的sdk有一个libuapi库是对这部分接口的封装，我在v851s上用DISP_LAYER_SET_CONFIG2会导致内核内存泄漏，你可以试试有没有同样的问题

我猜你遇到的问题可能和这段代码有关：

#if defined(CONF_KERNEL_IOMMU)
            ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_ENGINE_REL, 0);
            if (ret)
                logv("ENGINE_REL failed\n");

            struct user_iommu_param iommu_param;
            memset(&iommu_param, 0, sizeof(iommu_param));
            iommu_param.fd = tmp->fd_data.aw_fd;
            ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_FREE_IOMMU_ADDR, &iommu_param);
            if (ret)
                logv("FREE_IOMMU_ADDR err, ret %d\n", ret);
#endif

            /*close dma buffer fd*/
            close(tmp->fd_data.aw_fd);
            /* free buffer */
            handle_data.handle = tmp->fd_data.handle;

            ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_FREE, &handle_data);
            if (ret)
                logv("TON_IOC_FREE failed\n");

            aw_mem_list_del(&tmp->i_list);
            free(tmp);

            flag = 1;

        handle_data.handle = tmp->fd_data.handle;

内核版本为高版本（5.4）的时候，libuapi库在申请内存时根本没有初始化这个handle。然后释放内存的时候，却又用这个空的handle去调用释放的系统调用，内存根本不可能真正释放。换句话说，这个libuapi库只有在低版本内核上才有可能正常工作；高版本内核上，内存只能申请，无法释放。

laocai

@laocai 另外关于sunxi_display2.h文件，我在tina sdk的package中发现了七个路径下有这个文件，前面三个文件的内容各不相同，和内核使用的sunxi_display2.h文件也不同。
allwinner/libAWIspApi/src/isp522/libisp/include/device/sunxi_display2.h
allwinner/tina_multimedia/libcedarx/demo/xplayerdemo/sunxi_display2.h
allwinner/tina_multimedia/libcedarx/demo/recoderdemo/sunxi_display2.h
allwinner/libuapi/src/sunxi_display_v2.h
testtools/tinatest/testcase/base/production/displaytester/common/sunxi_display_v2.h
allwinner/tinacvr/src/middle_ware/sunxi_display_v2.h
gui/minigui/multimedia-test/src/middle_ware/sunxi_display_v2.h

比较了一下，和内核头文件最接近的是allwinner/libAWIspApi/src/isp522/libisp/include/device/sunxi_display2.h

allwinner/tinacvr/src/middle_ware/sunxi_display_v2.h和allwinner/libuapi/src/sunxi_display_v2.h比较接近，后者像是在前者基础上修改而来。

laocai

@laocai 如果要获得framebuffer的物理地址，就采用这个方法

void* sunxifb_mem_get_phyaddr(void *data) {
    if (data != NULL)
        return SunxiMemGetPhysicAddressCpu(memops, data);
    else
        return NULL;
}

laocai

@laocai 我找到了一个简单使用libuapi库的例子，在tina-package/blob/main/gui/littlevgl-8/lv_drivers/display/sunximem.c

/**
 * @file sunximem.c
 *
 */

/*********************
 *      INCLUDES
 *********************/
#include "sunximem.h"

#if USE_SUNXIFB_G2D

#include <stdio.h>
#include <ion_mem_alloc.h>

/*********************
 *      DEFINES
 *********************/

/**********************
 *      TYPEDEFS
 **********************/

/**********************
 *      STRUCTURES
 **********************/

/**********************
 *  STATIC PROTOTYPES
 **********************/
static struct SunxiMemOpsS *memops;

/**********************
 *  STATIC VARIABLES
 **********************/

/**********************
 *      MACROS
 **********************/

/**********************
 *   GLOBAL FUNCTIONS
 **********************/

bool sunxifb_mem_init() {
    if (memops)
        return true;

    memops = GetMemAdapterOpsS();
    if (SunxiMemOpen(memops) < 0) {
        perror("Error: cannot open ion device");
        return false;
    }

    return true;
}

void sunxifb_mem_deinit() {
    if (memops) {
        SunxiMemClose(memops);
        memops = NULL;
    }
}

void* sunxifb_mem_alloc(size_t size, char *label) {
    if (size == 0) {
        printf("size illegal.");
        return NULL;
    }

    void *alloc = SunxiMemPalloc(memops, size);
    if (alloc == NULL) {
        printf("couldn't allocate memory (%lu bytes).", (unsigned long) size);
        return NULL;
    }

#ifdef LV_USE_SUNXIFB_DEBUG
    printf("%s: sunxifb_mem_alloc=%p size=%lu bytes\n", label, alloc, (unsigned long) size);
#endif /* LV_USE_SUNXIFB_DEBUG */

    return alloc;
}

void sunxifb_mem_free(void **data, char *label) {
    if (*data != NULL) {
#ifdef LV_USE_SUNXIFB_DEBUG
        printf("%s: sunxifb_mem_free=%p\n", label, *data);
#endif /* LV_USE_SUNXIFB_DEBUG */
        SunxiMemPfree(memops, *data);
        *data = NULL;
    } else {
        printf("couldn't free memory.\n");
    }
}

void* sunxifb_mem_get_phyaddr(void *data) {
    if (data != NULL)
        return SunxiMemGetPhysicAddressCpu(memops, data);
    else
        return NULL;
}

void sunxifb_mem_flush_cache(void *data, size_t size) {
    SunxiMemFlushCache(memops, data, size);
}

/**********************
 *   STATIC FUNCTIONS
 **********************/

#endif

laocai

@laocai 而选择ion设备还是cedar设备，是由编译选项决定的。如果要采用DISP_LAYER_SET_CONFIG2，必须编译的时候选择SUNXI_ALLOC_IOMMU，不知道我的理解是否正确？我找到的关于全志disp的讨论

choice
prompt "Choose alloc method"
	default SUNXI_ALLOC_CMA
	help
	Select how to apply for memory.
config SUNXI_ALLOC_CMA
	bool "sunxi_alloc_cma"
config SUNXI_ALLOC_IOMMU
	bool "sunxi_alloc_iommu"
endchoice
choice
prompt "Choose layer alloc method"
	default LUAPI_LAYER_ALLOC_CHN
	help
	Select how to apply for layer.
config LUAPI_LAYER_ALLOC_CHN
	bool "luapi_layer_alloc_chn"
config LUAPI_LAYER_ALLOC_LAY
	bool "luapi_layer_alloc_lay"
endchoice
choice
prompt "Choose display rendering mode"
	default SUNXI_DISPLAY_DE
	help
		Choose display rendering mode, DE or GPU.
config SUNXI_DISPLAY_DE
	bool "DE"
config SUNXI_DISPLAY_GPU
	bool "GPU"
	help
	Define the macro SUNXI_DISPLAY_GPU, use GPU rendering, send the RGBA data
	after rendering to fb1 display, execute export FRAMEBUFFER=/dev/fb1
	before executing the program, no longer support DispSetEnable, DispSetSrcRect,
	DispSetZorder and DispSetIoctl functions.

	Define the macro SUNXI_DISPLAY_GPU and SUNXI_DISPLAY_GPU_BUF, use GPU rendering,
	send the RGBA data after rendering to layer. No longer support DispSetSrcRect functions.
	Default use.
endchoice
endif
endef

define Build/Prepare
	$(INSTALL_DIR) $(PKG_BUILD_DIR)/
	$(CP) ./src $(PKG_BUILD_DIR)/
endef

define Build/Configure
endef

ifeq ($(CONFIG_SUNXI_DISPLAY_TWO),y)
TARGET_CFLAGS+=-D__SUNXI_DISPLAY2__
TARGET_CFLAGS+=-D$(KERNEL_VERSION)
endif
ifeq ($(CONFIG_SUNXI_DISPLAY_ONE),y)
TARGET_CFLAGS+=-D__SUNXI_DISPLAY__
TARGET_CFLAGS+=-D$(KERNEL_VERSION)
endif

ifeq ($(CONFIG_SUNXI_ALLOC_CMA),y)
TARGET_CFLAGS+=-DCONF_KERNEL_CMA
endif
ifeq ($(CONFIG_SUNXI_ALLOC_IOMMU),y)
TARGET_CFLAGS+=-DCONF_KERNEL_IOMMU
endif

另外我还发现，linux 内核5.4版本的时候，libuapi库无法使用ion设备来分配framebuffer，原因就是代码里面根本没有实现alloc_data的赋值。

#if CONF_KERNEL_VERSION_OLD
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_ALLOC, &alloc_data);
    if (ret) {
        loge("ION_IOC_ALLOC error\n");
        goto ALLOC_OUT;
    }

    /* get dmabuf fd */
    fd_data.handle = alloc_data.handle;
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_MAP, &fd_data);
    if (ret) {
        loge("ION_IOC_MAP err, ret %d, dmabuf fd 0x%08x\n",
				ret, (unsigned int)fd_data.aw_fd);
        goto ALLOC_OUT;
    }
#else
    struct aw_ion_new_alloc_data data = {
        .len = alloc_data.aw_len,
        .heap_id_mask = alloc_data.aw_heap_id_mask,
        .flags = alloc_data.flags,
    };

    ret = ioctl(g_ion_alloc_context->fd, AW_ION_IOC_NEW_ALLOC, &data);
    if (ret < 0) {
        loge("AW_ION_IOC_NEW_ALLOC error\n");
        goto ALLOC_OUT;
    }
    fd_data.aw_fd = data.fd;
#endif

    /* get phy address */
#if defined(CONF_KERNEL_IOMMU)
    memset(&iommu_param, 0, sizeof(iommu_param));
    iommu_param.fd = fd_data.aw_fd;
    ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_ENGINE_REQ, 0);
    if (ret) {
        loge("ENGINE_REQ err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_GET_IOMMU_ADDR, &iommu_param);
    if (ret) {
        loge("GET_IOMMU_ADDR err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    addr_phy = iommu_param.iommu_addr;
#else
    memset(&phys_data, 0, sizeof(phys_data));
    phys_data.handle = alloc_data.handle;
    phys_data.size = size;
    custom_data.aw_cmd = ION_IOC_SUNXI_PHYS_ADDR;
    custom_data.aw_arg = (unsigned long)&phys_data;
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_CUSTOM, &custom_data);
    if (ret) {
        loge("ION_IOC_CUSTOM err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    addr_phy = phys_data.phys_addr;
#endif

laocai

@laocai 找到了libuapi\src\ion_alloc.c中的代码

/* return virtual address: 0 failed */
void* sunxi_ion_alloc_palloc(int size)
{
    aw_ion_allocation_info_t alloc_data;
    ion_fd_data_t fd_data;
    struct ion_handle_data handle_data;
    struct aw_ion_custom_info custom_data;
    sunxi_phys_data   phys_data;
    struct user_iommu_param iommu_param;

    int rest_size = 0;
    unsigned long addr_phy = 0;
    unsigned long addr_vir = 0;
    buffer_node * alloc_buffer = NULL;
    int ret = 0;
    char *cache_flag;
    pthread_mutex_lock(&g_ion_mutex_alloc);

    if (g_ion_alloc_context == NULL) {
		loge("call ion_alloc_open\n");
        goto ALLOC_OUT;
    }

    if (size <= 0) {
        loge("can not alloc size 0\n");
        goto ALLOC_OUT;
    }

    memset(&alloc_data, 0, sizeof(alloc_data));
    memset(&fd_data, 0, sizeof(fd_data));

    alloc_data.aw_len = (size_t)size;
    alloc_data.aw_align = ION_ALLOC_ALIGN ;

#if defined(CONF_KERNEL_IOMMU)
    alloc_data.aw_heap_id_mask = AW_ION_SYSTEM_HEAP_MASK | AW_ION_CARVEOUT_HEAP_MASK;
#else
    alloc_data.aw_heap_id_mask = AW_ION_DMA_HEAP_MASK | AW_ION_CARVEOUT_HEAP_MASK;
#endif
    cache_flag = getenv("MG_IONBUFFER");
    if(cache_flag == NULL || strcmp(cache_flag, "1") != 0){
        alloc_data.flags = AW_ION_CACHED_FLAG | AW_ION_CACHED_NEEDS_SYNC_FLAG;
    }

#if CONF_KERNEL_VERSION_OLD
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_ALLOC, &alloc_data);
    if (ret) {
        loge("ION_IOC_ALLOC error\n");
        goto ALLOC_OUT;
    }

    /* get dmabuf fd */
    fd_data.handle = alloc_data.handle;
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_MAP, &fd_data);
    if (ret) {
        loge("ION_IOC_MAP err, ret %d, dmabuf fd 0x%08x\n",
				ret, (unsigned int)fd_data.aw_fd);
        goto ALLOC_OUT;
    }
#else
    struct aw_ion_new_alloc_data data = {
        .len = alloc_data.aw_len,
        .heap_id_mask = alloc_data.aw_heap_id_mask,
        .flags = alloc_data.flags,
    };

    ret = ioctl(g_ion_alloc_context->fd, AW_ION_IOC_NEW_ALLOC, &data);
    if (ret < 0) {
        loge("AW_ION_IOC_NEW_ALLOC error\n");
        goto ALLOC_OUT;
    }
    fd_data.aw_fd = data.fd;
#endif

    /* mmap to user */
    addr_vir = (unsigned long)mmap(NULL, alloc_data.aw_len, \
        PROT_READ|PROT_WRITE, MAP_SHARED, fd_data.aw_fd, 0);

    if ((unsigned long)MAP_FAILED == addr_vir) {
        loge("mmap err, ret %u\n", (unsigned int)addr_vir);
        addr_vir = 0;
        goto ALLOC_OUT;
    }

    /* get phy address */
#if defined(CONF_KERNEL_IOMMU)
    memset(&iommu_param, 0, sizeof(iommu_param));
    iommu_param.fd = fd_data.aw_fd;
    ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_ENGINE_REQ, 0);
    if (ret) {
        loge("ENGINE_REQ err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_GET_IOMMU_ADDR, &iommu_param);
    if (ret) {
        loge("GET_IOMMU_ADDR err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    addr_phy = iommu_param.iommu_addr;
#else
    memset(&phys_data, 0, sizeof(phys_data));
    phys_data.handle = alloc_data.handle;
    phys_data.size = size;
    custom_data.aw_cmd = ION_IOC_SUNXI_PHYS_ADDR;
    custom_data.aw_arg = (unsigned long)&phys_data;
    ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_CUSTOM, &custom_data);
    if (ret) {
        loge("ION_IOC_CUSTOM err, ret %d\n", ret);
        addr_phy = 0;
        addr_vir = 0;
        goto ALLOC_OUT;
    }
    addr_phy = phys_data.phys_addr;
#endif

    alloc_buffer = (buffer_node *)malloc(sizeof(buffer_node));
    if (alloc_buffer == NULL) {
        loge("malloc buffer node failed");

        /* unmmap */
        ret = munmap((void*)addr_vir, alloc_data.aw_len);
        if (ret) {
            loge("munmap err, ret %d\n", ret);
        }

#if defined(CONF_KERNEL_IOMMU)
        ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_ENGINE_REL, 0);
        if (ret)
            logv("ENGINE_REL failed\n");

        ret = ioctl(g_ion_alloc_context->cedar_fd, AW_MEM_FREE_IOMMU_ADDR, &iommu_param);
        if (ret)
            logv("FREE_IOMMU_ADDR err, ret %d\n", ret);
#endif

        /* close dmabuf fd */
        close(fd_data.aw_fd);

        /* free buffer */
        handle_data.handle = alloc_data.handle;
        ret = ioctl(g_ion_alloc_context->fd, AW_MEM_ION_IOC_FREE, &handle_data);
        if(ret)
            loge("ION_IOC_FREE err, ret %d\n", ret);

        addr_phy = 0;
        addr_vir = 0; /* value of MAP_FAILED is -1, should return 0*/

        goto ALLOC_OUT;
    }
    alloc_buffer->phy = addr_phy;
    alloc_buffer->vir = addr_vir;
    alloc_buffer->user_virt = addr_vir;
    alloc_buffer->size = size;
    alloc_buffer->fd_data.handle = fd_data.handle;
    alloc_buffer->fd_data.aw_fd = fd_data.aw_fd;

    aw_mem_list_add_tail(&alloc_buffer->i_list, &g_ion_alloc_context->list);

#if DEBUG_ION_REF==1
    cdx_use_mem += size;
    int i = 0;
    for (i = 0; i < ION_BUF_LEN; i++) {
        if (ion_buf_nodes_test[i].addr == 0
				&& ion_buf_nodes_test[i].size == 0) {
            ion_buf_nodes_test[i].addr = addr_vir;
            ion_buf_nodes_test[i].size = size;
            break;
        }
    }

    if (i>= ION_BUF_LEN) {
        loge("error, ion buf len is large than [%d]", ION_BUF_LEN);
    }
#endif

ALLOC_OUT:
    pthread_mutex_unlock(&g_ion_mutex_alloc);
    return (void*)addr_vir;
}

根据配置不同，libuapi使用ion设备或者cedar设备来分配内存。

laocai

@zheng1 之前我只用了tina linux内核，没有安装tina sdk。
等我把tina sdk安装了看看，全志要能提供更详细的使用说明就好了。

laocai

另外全志的文档中提到sunxi_display2.h，我发现这个头文件不能直接用于应用程序开发。原因是它引用了很多只有内核中才用到的c语言类型，如s32等。请问全志的sunxi应用程序开发到底要使用什么样的开发环境呢？屏幕截图 2023-12-01 174857.png

laocai

@xiaowenge 图片没挂，只是不支持外链。

laocai

@laocai

这是我网上搜到的一张图片，这段代码使用了android的ion库来创建framebuffer。sunxi图层创建必须要用ion库吗？

PS，这段代码是采用DISP_LAYER_SET_CONFIG2配置，所以是用ion memory，那么DISP_LAYER_SET_CONFIG配置呢，DISP_LAYER_SET_CONFIG配置的时候，应用程序该如何准备图层的framebuffer?有没有大神能解释一下

laocai

通过ioctl函数传递的图层参数，framebuffer地址（即下面的add[0]）到底是物理地址还是虚拟地址？这个framebuffer由应用层提供，有没有什么特别要求？比如地址对齐要求、连续内存要求等等。另外哪里能找到说明文档或者参考代码？

/* 选择图层所需的通道以及本图层参数 */
config.channel = 0;
config.layer_id = 0;
config.info.enable = 1;
config.info.mode = LAYER_MODE_BUFFER;
config.info.fb.addr[0] = (unsigned long long)mem_in; //FB物理地址
config.info.fb.size[0].width = width;
config.info.fb.size[0].height = height;
config.info.fb.align[0] = 4; //4bytes align
config.info.fb.format = DISP_FORMAT_ARGB_8888;

/* 上层调用DE显示引擎所用的ioctl接口 */
arg[0] = 0; //screen_id 0表示选择通路0
arg[1] = (unsigned long)&config;
arg[2] = 1;//表示设置图层数量为1
ret = ioctl(disphd, DISP_LAYER_SET_CONFIG, (void *)arg);

laocai

@houxiaoni 全志对gstreamer做了适配支持，怎么tina的开发文档里都没有提及？还是我漏看了什么重要内容？我看到的文档，多媒体开发只有tplayer工具这一种，楼主能否澄清一下？

laocai

@xiaowenge 在线申请，还是要联系sales？

laocai

@yuzukitsuru V3S的tina linux sdk，需要单独向全志申请，还是已经公开？

laocai

规划中一个产品，主要用gstreamer做h264硬解，然后显示到lcd上。是基于tina linux开发好，还是主线linux开发好？要求就是资料多，坑少一点。能够快一点做出来。
另外，这个产品规划生命周期5年，总产量1000台。请问全志会为这样的产品提供技术支持吗？

laocai 发布的帖子