Re: [PATCH 28/28] drm/xe/pf: Refactor strlcat() to seq_buf in sched_group_engines_read()
From: Kees Cook <kees@kernel.org>
Date: 2026-09-15 18:30:27
Also in:
dri-devel, intel-wired-lan, intel-xe, linux-devicetree, linux-edac, linux-gpio, linux-hardening, linux-input, linux-media, linux-mips, linux-nfs, linux-samsung-soc, linux-scsi, linux-sound, linux-wireless, lkml, loongarch
On Tue, Sep 15, 2026 at 08:18:45AM +0000, Bill Wendling wrote:
quoted hunk ↗ jump to hunk
diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c b/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c
This is the same file touched as https://lore.kernel.org/all/20260915081910.4142719-28-morbo@google.com/ (local) and doing the same thing, roughly? I think it'd be best to combine these, especially since one depends on the other's addition of seq_buf.h
quoted hunk ↗ jump to hunk
index 5db7fc067c4d..548a50f74422 100644--- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c@@ -377,8 +377,10 @@ static ssize_t sched_group_engines_read(struct file *file, char __user *buf, struct xe_hw_engine *hwe; enum xe_hw_engine_id id; char engines[128]; + struct seq_buf s; + const char *s_str; - engines[0] = '\0'; + seq_buf_init(&s, engines, sizeof(engines));
As with the others, DECLARE_SEQ_BUF should be used.
quoted hunk ↗ jump to hunk
if (group < num_groups) { for_each_hw_engine(hwe, gt, id) {@@ -386,15 +388,14 @@ static ssize_t sched_group_engines_read(struct file *file, char __user *buf, u16 guc_logical_instance = xe_hwe_guc_logical_instance(hwe); u32 mask = groups[group].engines[guc_class]; - if (mask & BIT(guc_logical_instance)) { - strlcat(engines, hwe->name, sizeof(engines)); - strlcat(engines, " ", sizeof(engines)); - } + if (mask & BIT(guc_logical_instance)) + seq_buf_printf(&s, "%s ", hwe->name); } - strlcat(engines, "\n", sizeof(engines)); + seq_buf_puts(&s, "\n"); } - return simple_read_from_buffer(buf, count, ppos, engines, strlen(engines)); + s_str = seq_buf_str(&s); + return simple_read_from_buffer(buf, count, ppos, s_str, strlen(s_str));
While this is the correct length, it does illustrate why I think we need a new seq_buf helper to get the strlen result detected from NUL termination like seq_buf_str(). -Kees -- Kees Cook