<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40"> <head> <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1"> <meta name="Generator" content="Microsoft Word 15 (filtered medium)"> <style></style> </head> <body lang="EN-US" link="blue" vlink="purple" style="word-wrap:break-word"> [AMD Official Use Only] <div class="WordSection1"> <o:p> </o:p> <div> On 2/21/2022 7:58 PM, Christian König wrote:<o:p></o:p> </div> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Am 21.02.22 um 15:19 schrieb Somalapuram, Amaranath: <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> [AMD Official Use Only] On 2/21/2022 7:09 PM, Christian König wrote: <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Am 21.02.22 um 14:34 schrieb Somalapuram Amaranath: <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> List of register populated for dump collection during the GPU reset. Signed-off-by: Somalapuram Amaranath <a href="mailto:Amaranath.Somalapuram@amd.com"> <Amaranath.Somalapuram@amd.com></a> --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 96 +++++++++++++++++++++ 2 files changed, 100 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index b85b67a88a3d..6e35f2c4c869 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1097,6 +1097,10 @@ struct amdgpu_device { struct amdgpu_reset_control *reset_cntl; uint32_t ip_versions[HW_ID_MAX][HWIP_MAX_INSTANCE]; + + /* reset dump register */ + uint32_t *reset_dump_reg_list; + int num_regs; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 164d6a9e9fbb..69c0a28deeac 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -1609,6 +1609,100 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL, DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL, amdgpu_debugfs_sclk_set, "%llu\n"); +static ssize_t amdgpu_reset_dump_register_list_read(struct file *f, + char __user *buf, size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; + char reg_offset[11]; + int i, ret, len = 0; + + if (*pos) + return 0; + + ret = down_read_killable(&adev->reset_sem); + + if (ret) + return ret; <o:p></o:p> </blockquote> We usually don't have an empty line between function call and checking the return code. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + + for (i = 0; i < adev->num_regs; i++) { + down_read(&adev->reset_sem); <o:p></o:p> </blockquote> That here will just crash because we have already locked the semaphore before the loop. <o:p></o:p> </blockquote> unfortunately it did not crash. Sorry I misunderstood your earlier comments. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + sprintf(reg_offset, "0x%x ", adev->reset_dump_reg_list[i]); + up_read(&adev->reset_sem); + ret = copy_to_user(buf + len, reg_offset, strlen(reg_offset)); + + if (ret) + goto error; + + len += strlen(reg_offset); <o:p></o:p> </blockquote> And here the down_read_killable() is missing. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + } + + up_read(&adev->reset_sem); + ret = copy_to_user(buf + len, "\n", 1); + + if (ret) + return -EFAULT; + + len++; + *pos += len; + + return len; +error: + up_read(&adev->reset_sem); + return -EFAULT; +} + +static ssize_t amdgpu_reset_dump_register_list_write(struct file *f, + const char __user *buf, size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; + char *reg_offset, *reg, reg_temp[11]; + uint32_t *tmp; + int ret, i = 0, len = 0; + + do { + reg_offset = reg_temp; <o:p></o:p> </blockquote> I think you can just drop the reg_offset variable; <o:p></o:p> </blockquote> strsep takes only pointer as input, this is workaround. <o:p> </o:p> </blockquote> Ah, now I see what you are doing here. Please don't do it like that. Better use memchr() instead. <o:p></o:p> </blockquote> memchr will not work. I couldn't find any other string API I can use.<o:p></o:p> other references similar to code:<o:p></o:p> <div> <div> static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)<o:p></o:p> </div> <div> {<o:p></o:p> </div> <div> adev->enable_virtual_display = false;<o:p></o:p> </div> <o:p> </o:p> <div> if (amdgpu_virtual_display) {<o:p></o:p> </div> <div> const char *pci_address_name = pci_name(adev->pdev);<o:p></o:p> </div> <div> char *pciaddstr, *pciaddstr_tmp, *pciaddname_tmp, *pciaddname;<o:p></o:p> </div> <o:p> </o:p> <div> pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL);<o:p></o:p> </div> <div> pciaddstr_tmp = pciaddstr;<o:p></o:p> </div> <div> while ((pciaddname_tmp = strsep(&pciaddstr_tmp, ";"))) {<o:p></o:p> </div> <div> pciaddname = strsep(&pciaddname_tmp, ",");<o:p></o:p> </div> </div> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + memset(reg_offset, 0, 11); + ret = copy_from_user(reg_offset, buf + len, min(11, ((int)size-len))); + + if (ret) + goto error_free; + + reg = strsep(&reg_offset, " "); + tmp = krealloc_array(tmp, 1, sizeof(uint32_t), GFP_KERNEL); <o:p></o:p> </blockquote> That must be krealloc_array(tmp, i, ... not krealloc_array(tmp, 1, ... ! <o:p></o:p> </blockquote> I thought it will append (if not it should have crashed or some kernel dump) <o:p></o:p> </blockquote> No, krealloc_array works similar to realloc() in userspace. You need to give it the full size of the necessary space. Regards, Christian. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Regards, Christian. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + ret = kstrtouint(reg, 16, &tmp[i]); + + if (ret) + goto error_free; + + len += strlen(reg) + 1; + i++; + + } while (len < size); + + ret = down_write_killable(&adev->reset_sem); + + if (ret) + goto error_free; + + swap(adev->reset_dump_reg_list, tmp); + adev->num_regs = i; + up_write(&adev->reset_sem); + ret = size; + +error_free: + kfree(tmp); + return ret; +} + + + +static const struct file_operations amdgpu_reset_dump_register_list = { + .owner = THIS_MODULE, + .read = amdgpu_reset_dump_register_list_read, + .write = amdgpu_reset_dump_register_list_write, + .llseek = default_llseek +}; + int amdgpu_debugfs_init(struct amdgpu_device *adev) { struct dentry *root = adev_to_drm(adev)->primary->debugfs_root; @@ -1672,6 +1766,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) &amdgpu_debugfs_test_ib_fops); debugfs_create_file("amdgpu_vm_info", 0444, root, adev, &amdgpu_debugfs_vm_info_fops); + debugfs_create_file("amdgpu_reset_dump_register_list", 0644, root, adev, + &amdgpu_reset_dump_register_list); adev->debugfs_vbios_blob.data = adev->bios; adev->debugfs_vbios_blob.size = adev->bios_size; <o:p></o:p> </blockquote> </blockquote> > <o:p></o:p> </blockquote> <o:p> </o:p> </blockquote> </div> </body> </html>