fix(scan): require sudo access for dmesg scanning (#13)

* fix(scan): require sudo access for dmesg scanning Signed-off-by: Gyuho Lee <[email protected]> * update Signed-off-by: Gyuho Lee <[email protected]> --------- Signed-off-by: Gyuho Lee <[email protected]>
leptonai · Aug 20, 2024 · 1ddef0a · 1ddef0a
1 parent 251c654
commit 1ddef0a
Show file tree

Hide file tree

Showing 4 changed files with 17 additions and 3 deletions.
diff --git a/cmd/gpud/command/diagnose.go b/cmd/gpud/command/diagnose.go
@@ -13,7 +13,7 @@ import (
 
 func cmdDiagnose(cliContext *cli.Context) error {
 	if os.Geteuid() != 0 {
-		return errors.New("diagnose requires root")
+		return errors.New("requires sudo/root access to diagnose GPU issues")
 	}
 
 	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)

diff --git a/components/accelerator/nvidia/query/nvidia_peermem.go b/components/accelerator/nvidia/query/nvidia_peermem.go
@@ -12,7 +12,7 @@ const peerMemModule = "nvidia_peermem"
 
 func CheckLsmodPeermemModule(ctx context.Context) (*LsmodPeermemModuleOutput, error) {
 	if os.Geteuid() != 0 {
-		return nil, errors.New("nvidia_peermem check requires root")
+		return nil, errors.New("requires sudo/root access to check if ib_core is using nvidia_peermem")
 	}
 
 	b, err := exec.CommandContext(ctx, "sudo", "lsmod").CombinedOutput()

diff --git a/components/accelerator/nvidia/query/query.go b/components/accelerator/nvidia/query/query.go
@@ -298,6 +298,10 @@ func (o *Output) PrintInfo(debug bool) {
 	}
 
 	if o.SMI != nil {
+		if len(o.SMI.GPUs) > 0 {
+			fmt.Printf("%s product name: %s (nvidia-smi)\n", checkMark, o.SMI.GPUs[0].ProductName)
+		}
+
 		if errs := o.SMI.FindGPUErrs(); len(errs) > 0 {
 			fmt.Printf("%s scanned nvidia-smi -- found %d error(s)\n", warningSign, len(errs))
 			for _, err := range errs {
@@ -356,6 +360,10 @@ func (o *Output) PrintInfo(debug bool) {
 	}
 
 	if o.NVML != nil {
+		if len(o.NVML.DeviceInfos) > 0 {
+			fmt.Printf("%s name: %s (NVML)\n", checkMark, o.NVML.DeviceInfos[0].Name)
+		}
+
 		for _, dev := range o.NVML.DeviceInfos {
 			fmt.Printf("\n\n##################\nNVML scan results for %s\n\n", dev.UUID)
 

diff --git a/components/diagnose/scan.go b/components/diagnose/scan.go
@@ -2,7 +2,9 @@ package diagnose
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"os"
 	"time"
 
 	nvidia_query "github.com/leptonai/gpud/components/accelerator/nvidia/query"
@@ -23,6 +25,10 @@ const (
 
 // Runs the scan operations.
 func Scan(ctx context.Context, lines int, debug bool) error {
+	if os.Geteuid() != 0 {
+		return errors.New("requires sudo/root access in order to scan dmesg errors")
+	}
+
 	fmt.Printf("\n\n%s scanning the host\n\n", inProgress)
 
 	if nvidia_query.SMIExists() {
@@ -45,7 +51,7 @@ func Scan(ctx context.Context, lines int, debug bool) error {
 			} else {
 				output.PrintInfo(debug)
 
-				fmt.Printf("%s checking nvidia xid errors\n", inProgress)
+				fmt.Printf("\n%s checking nvidia xid errors\n", inProgress)
 
 				select {
 				case <-ctx.Done():